VOLUME 14, ISSUE 6, JUNE 2025

Abstract: Recent analysis have identified a significant rise in transactional fraud, where bad actors seek to deceive individuals or firms into unauthorized financial actions. Traditional fraud detection systems frequently struggle to effectively identify such activities, leading to financial damages and security breaches. Addressing this problem requires employing sophisticated machine learning techniques specially designed to detect transactional fraud. This study presents a novel fraud detection method called “filter”, aimed at uncovering misleading transactional behaviours. By employing tailored features to reveal fraudulent patterns and activities, our filter achieves a remarkable accuracy of over 99.01% in distinguishing fraudulent transactions from legitimate ones, while maintaining a low false positive rate. Our approach was evaluated with a dataset comprising 746 instances of fraudulent transactions and 4822 instances of legitimate transactions. The results underscore the superior performance of our filter compared to existing methods, particularly in accurately detecting fraudulent transactions. Moreover, our hybrid NB-ANN model achieves the highest accuracy of 99.01%, outperforming both Naïve Bayes (98.57%) and Artificial Neural Network (98.12%) techniques. This highlights the effectiveness of the hybrid method in boosting detection accuracy for transactional fraud. Implementing our filter and leveraging the hybrid NB-ANN model, organizations can greatly improve their ability to detect and prevent fraudulent activities thereby protecting their financial assets and maintaining customer trust.

Keywords: Machine learning, Predictive model, transaction fraud, dataset, hybrid NB-ANN, Filter, Legitimate.

Abstract: Motorcycle riders face heightened risks due to limited visibility and lack of structural protection, with blind spot detection being a critical safety challenge. This project presents a Motorcycle Blind Spot Detection System that utilizes both ultrasonic and radar sensors to enhance rider awareness and reduce accidents. Ultrasonic sensors provide precise short-range detection, while radar sensors ensure reliable tracking of moving objects in wider ranges and varying weather conditions. Data from both sensors is processed by a microcontroller to identify potential threats in blind spot zones. Upon detection, the system activates visual alerts via LED indicators, with provisions for future integration of auditory and haptic feedback. The dual-sensor fusion approach increases detection accuracy and minimizes false positives, offering a cost-effective, scalable, and real-time solution to improve motorcycle safety.

Keywords: Motorcycle safety, Blind spot detection, Sensor fusion, Real-time alert system, Rider awareness, Collision prevention.

Abstract: The Telemedicine System for Access to Healthcare is created to mitigate the common barriers that affect most people's ability to seek health services. Long distance, vast distance, high cost, mobility, etc. Many people's factors prevent them from accessing the medical services required, which may help interventions in remote consultations, diagnosis, and monitoring. Timely medical attention is the ultimate goal of the initiative, especially for the disadvantaged communities. The other side of the technology is to improve healthcare delivery. Automated scheduling and AI assisted diagnostics are just some of the smart features that have been incorporated into the system to personalize services according to an individual's needs. The program does more than just providing healthcare access; it has that potential to create a very positive social impact. It will facilitate people's individual and collective vulnerable populations such as rural area people, especially in availing the required health care. This will also prove beneficial in reducing the number of people who go to hospitals, thereby averting congestion and carbon emissions in addition to promoting active health management. Ultimately this leads to healthier communities and enhances the quality of life.

Keywords: Telemedicine, Remote consultations, underserved population, Automated scheduling, Healthcare efficiency.

Abstract: The dual responsibility of managing professional and domestic roles often places married women teachers under considerable stress, which can significantly affect their mental health. This study investigates the key stress factors contributing to psychological distress among married women educators in India and examines their impact on mental well-being. The teaching profession, while considered noble and fulfilling, is increasingly becoming a source of stress due to work overload, administrative pressure, lack of institutional support, and ever-growing expectations. When combined with domestic responsibilities such as childcare, household chores, and societal obligations, the cumulative burden becomes immense for married women teachers. The study employed a descriptive survey design, utilizing standardized stress and mental health assessment tools among a purposive sample of 200 married women teachers from primary, secondary, and higher education institutions. Data were analyzed using descriptive statistics and correlation analysis to determine the relationship between identified stressors and mental health indicators such as anxiety, depression, and emotional exhaustion. Findings reveal that occupational stressors (e.g., time constraints, student behavior, administrative tasks), familial obligations, and societal expectations were significant contributors to mental distress. A strong positive correlation was found between stress levels and symptoms of poor mental health, suggesting the urgent need for institutional and community-level interventions. The study emphasizes the importance of mental health support systems, flexible work policies, and counseling services in educational settings. The research offers valuable insights into the often-overlooked psychosocial challenges of married women educators, urging stakeholders to adopt a holistic approach in promoting their mental well-being. These findings are critical for educational policymakers, school administrators, and mental health professionals working toward gender-sensitive reforms.

Keywords: Married women teachers, stress factors, mental health, occupational stress, work-life balance, emotional well-being, psychological distress, India, gender roles, educational sector

Abstract: Mental health issues such as stress, anxiety, and depression are increasingly prevalent in modern society, adversely affecting individuals’ quality of life and productivity. Yoga, a holistic mind-body practice with origins in ancient India, combines physical postures, controlled breathing, and meditation to promote overall health and well-being. This study explores the mental health status of adults who regularly practice yoga compared to non-practitioners. Using standardized psychological measures including the Depression Anxiety Stress Scale (DASS-21) and General Health Questionnaire (GHQ-12), mental health parameters of 175 participants (75 regular yogic practitioners and 75 non-practitioners) were assessed. The study found that regular yogic practitioners exhibited significantly lower levels of depression, anxiety, and stress, along with improved psychological well-being compared to non-practitioners. These findings support existing literature on yoga’s positive effects on mental health and suggest that integrating yoga into daily routines can be an effective complementary approach to managing psychological distress. The study also discusses potential mechanisms behind yoga’s benefits, including regulation of the autonomic nervous system, reduction of cortisol levels, and enhancement of mindfulness and emotional regulation. Limitations such as sample size and self-report measures are acknowledged, with recommendations for future longitudinal and experimental research. This research reinforces yoga’s role as a valuable tool in promoting mental wellness in diverse populations, highlighting its potential for inclusion in mental health promotion and prevention programs.

Keywords: Yoga, Mental Health, Stress Reduction, Anxiety, Psychological Well-being

Keywords: Gamification, Natural Language Processing (NLP), Large Language Models (LLMs), mental Health, Emotional Health.

Abstract: Pharm Assist The issue of violence against women continues to pose a serious global challenge, underscoring the need for accessible digital platforms that foster awareness, offer support, and enable timely intervention. Mahila Suraksha Nyayavani is an integrated crime reporting website specifically designed to empower women by offering vital information, educational content, and direct access to law enforcement services. The platform addresses three key categories of violence—digital, emotional, and physical—and presents informative articles that highlight actual case studies and preventive strategies. It also provides instructional self-defense videos aimed at improving individual safety, along with a streamlined reporting form that allows victims to contact the police directly. To enhance accessibility, the platform features an AI-enabled chatbot that communicates in Kunglish (a hybrid of Kannada and English), guiding users through the website, assisting with incident reporting, and offering immediate support. This paper explores the platform’s multi-dimensional strategy by analyzing user interactions, collected feedback, and its potential to enhance the efficiency and responsiveness of digital crime reporting mechanisms.

Keywords: Digital Crime Reporting, Women’s Empowerment, Safety Education, AI-Based Chatbot Support, Kunglish Interaction, Legal Aid for Women, Technology in Law Enforcement, Emergency Communication, Police Integration Tools.

Abstract: Plants play a crucial role in food production, but diseases threaten crop yields and quality. Traditional manual inspection is time-consuming and inconsistent, while AI-powered detection offers a faster, more reliable solution. Using deep learning and CNNs, this system analyzes images of leaves, stems, and roots to classify diseases accurately. It also features an interactive chatbot to assist farmers with symptoms, treatments, and prevention. This paper explores advancements in AI-driven plant disease detection, evaluates its performance using a Kaggle dataset, and discusses challenges like dataset diversity and computing power. Future improvements aim to enhance multilingual support and accessibility for farmers. Grapes, a commercially significant crop, are highly vulnerable to leaf, stem, and fruit diseases. Early detection is essential for protecting yields. This project proposes a CNN-based deep learning approach to identify grape leaf diseases with high accuracy, reducing manual effort and providing timely decision support. Keyword: Plant disease detection, Deep learning, CNNs, AI in agriculture, Crop health monitoring, Grape leaf disease classification, Automated disease diagnosis, Image recognition in agriculture, smart farming solutions.

Abstract: This paper presents an Ayurvedic healthcare system powered by AI that focuses on the early detection and prevention of cancer through the analysis of individual body constitution (dosha). The system utilizes classic Ayurvedic wisdom and contemporary machine learning models to categorize users' doshas—Vata, Pitta, and Kapha—according to health input information. It makes personalized suggestions regarding diet, lifestyle, and yoga to achieve doshic equilibrium. One of the major innovations is the ability to identify patterns of dosha imbalances that would signify precursor signs of cancer. The platform incorporates support for wearable devices, allowing real-time monitoring of health through Google Fit and Apple Health APIs. The hybrid solution offers preventive healthcare as well as wellness support, integrating ancient healing traditions with present-day AI technologies. Our findings have encouraging potential for real-life applications in integral cancer risk analysis and prevention based on lifestyle.

Abstract: The allure of substantial returns in the stock market attracts countless investors, but the inherent volatility of stock prices—shaped by numerous dynamic factors—poses significant risks. To mitigate these uncertainties, investors often rely on analytical methods. One of the most pressing challenges in this domain is the accurate prediction of stock prices, making financial time series forecasting a key area where machine learning demonstrates immense potential. Research highlights that sophisticated forecasting techniques can effectively anticipate market trends. This study harnesses the capabilities of big data through the Apache Spark framework, enabling real-time analysis of stock trading volumes via a well-structured trading volume index. The system is designed to issue risk alerts corresponding to varying trading volume thresholds, thus empowering investors with timely and insightful data for improved decision-making. The results indicate that investors operating in volatile markets can enhance their financial outcomes by leveraging trading volume-based risk assessments. To support this, the study employs foundational machine learning algorithms such as linear regression and random forest for risk prediction related to stock performance.

Keywords: Computer Vision, Linear Regression, Machine Learning, Data Analytics, Risk Assessment, Portfolio optimization

Abstract: Job interview preparation is crucial, but many candidates find it difficult. Candidates often lack interactivity or real-time feedback when using traditional study methods. We introduce an AI-driven mock interview assistant with an animated 2D avatar and voice interaction to simulate realistic interview scenarios. Our system employs advanced natural language processing (OpenAI GPT API) for dynamic question generation and answer review, automated voice input recognition (Google Speech-to-Text) to receive user answers, and speech output as well as real-time lip-sync (Live2D Cubism and Wav2Lip) to animate an interviewer. The assistant combines these capabilities can conduct technical and HR mock interviews, critique answers, and engage users through auditory, visual, and textual communication. We describe the system design, AI and animation techniques, and potential applications in interview practice, online education, and interactive training. Our experiments show that an embodied conversational agent can improve the interview practice experience by adding interactivity, realism, and engagement.

Keywords: Mock Interview, Conversational AI, 2D Animated Avatar, Speech Recognition, Lip Sync, Natural Language Processing, Interview Preparation, Virtual Assistant, EdTech, Live2D Cubism, OpenAI GPT, Wav2Lip.

Abstract: Fruit Analysis through image processing is a method applied for detecting the detection of fruits by a given algorithm. This project is also applied in detecting the defected fruit among a set of fruits by the image processing method. The aim of this task is to avoid health risks by consuming that defected fruit. This processing method goes through various steps for the classification and forecasting of the defective fruit. This paper provides a comprehensive overview of different techniques i.e., preprocessing, segmentation, feature extraction, a designation which termed fruits and vegetable quality in terms of color, texture, size, shape, and defects. In this paper, a comparative analysis of another algorithm suggested by researchers for the quality testing of fruits has been conducted.

Keywords: Image processing, Machine Learning, Deep Learning, CNN, Decision Tree Classifier.

Abstract: In today's digital world, organizations rely on distributed storage systems to manage vast amounts of data across multiple servers. Each server, or host, is responsible for storing a specific portion of the data and takes backups at different time intervals to ensure reliability and disaster recovery. However, these backups are not always synchronized, meaning that when a system failure occurs, restoring data from different recovery points can lead to inconsistencies. This can cause issues like missing or outdated information, transactional mismatches, and operational disruptions. To solve this challenge, we propose an intelligent recovery point selection method that ensures the most consistent restoration of data. Our algorithm, inspired by the A* search technique, systematically evaluates all possible backup combinations and selects the set that minimizes the time difference across all hosts. By using an optimized heap-based selection process, it efficiently finds the most synchronized recovery points, reducing data inconsistency and improving reliability. Unlike traditional recovery methods that rely on manual selection or simple rules, our approach is automated, scalable, and computationally efficient. It can be applied in industries such as cloud computing, finance, healthcare, and e-commerce, where maintaining accurate and consistent data is critical. In the future, our solution can be further enhanced with machine learning to predict failures and optimize recovery strategies.

Keywords: Distributed storage systems, data consistency, backup recovery, asynchronous backups, recovery point selection, A* search algorithm, data integrity, system failure, optimized restoration, machine learning, disaster recovery, cloud computing.

Abstract: In modern digital systems, ensuring data integrity and identity verification has become vital, particularly in areas like government records, legal documents, and banking. This paper presents a survey of a biometric watermarking system that combines iris and fingerprint traits to enhance authentication. The proposed approach uses a Rubik Cube- based encryption algorithm to secure extracted biometric features, which are then verified using Convolutional Neural Networks (CNNs). The objective is to embed the encrypted biometric information into host images, producing a watermark that is resistant to tampering and forgery. This paper reviews the underlying techniques, compares them with existing methods, and highlights the security and privacy advantages of the system. The study is further motivated by recent findings on the privacy risks of facial recognition systems, reinforcing the need for multi-modal, secure biometric authentication.

Keywords: Biometric Watermarking, Iris Recognition, Fingerprint Authentication, Rubik Encryption, CNN, Image Security, Multi-Modal Biometrics.

Abstract: Legacy systems often suffer from a lack of clear ownership of code files and documents, leading to significant challenges in maintenance, security, and operational stability. Without defined accountability, it becomes difficult to track changes, address bugs, and implement necessary updates, increasing the risk of security vulnerabilities due to outdated or deprecated components. This project introduces a systematic approach for identifying unmonitored legacy data, categorizing unmaintained files, and establishing an efficient management framework. The proposed solution leverages automated techniques to analyze file metadata, assess modification patterns, and determine ownership attribution. Additionally, the system generates reports on unmaintained code and unmanaged documents, facilitating informed decision-making for risk mitigation. By implementing structured file ownership and maintenance protocols, organizations can enhance system security, improve operational efficiency, and ensure long-term software sustainability.

Keywords: Unmonitored, Legacy Data, Metadata, Unmaintained Files.

Abstract: The systematic case paper management system for homeopathy practitioners streamlines the administrative tasks within homeopathic hospitals, enabling efficient handling and reporting of patient data. This system allows doctors to easily add, edit, search, view, and print case papers, as well as record follow-up details for each patient. Additionally, it provides comprehensive visualizations, including graphs and pie charts, to track patient data by various metrics such as total case papers, monthly and weekly case papers, and age distribution. This project enhances operational efficiency, offering quick access to vital information and facilitating informed decision-making in patient care.

Keywords: Homeopathy, Case Paper Management, Medical Software, Patient Record System, Healthcare Automation, Homeopathic Practice Management, Digital Health Records, Patient Data Visualization, Medical Reporting Tools, Healthcare Technology

Abstract: This research delves into the development of digital art through the blending of artificial intelligence (AI), gesture detection, and automated precision software, and it does this in a futuristic manner through human- computer creativity. With conventional art converging with smart technology, this research looks at how artists are able to use hand movements to intuitively engage digital canvases, and AI enhances the work by fine- tuning shapes and composition in real-time. Through the use of computer vision and machine learning techniques, the system converts expressive human movement into organized yet tailored artwork, retaining the emotional richness of handwork together with the accuracy of electronic tools. Through qualitative analyses, such as user testing and prototyping, the study demotes enhancements in creative flow, ease of use for non-artists, and the future of collaborative human-AI co-creation. The research is in line with Human- Centered Design and Technological Augmentation theories, finally suggesting that AI can be a helper and creative collaborator in the creation of the next form of art.

Keywords: AI art, gesture recognition, digital painting, shape automation, human-AI collaboration, creative technology, computer vision, generative design.

Abstract: As cloud-based applications continue to grow in popularity, safeguarding virtual network access has emerged as a major cybersecurity concern. This study introduces a Cloud-Based Virtual Network Traffic Monitoring Framework that strengthens security by tracking and evaluating both inbound and outbound traffic within a web application environment. The system records comprehensive traffic logs during each user login session and securely stores them in the cloud infrastructure. To detect unauthorized access, it leverages a combination of machine learning models: autoencoders for unsupervised pattern recognition and logistic regression for supervised classification. This dual-model strategy enables the system to effectively understand typical access behaviors and flag anomalies. Upon identifying an unauthorized IP address, the system blocks further access attempts from that source in real time. By automating access control and anomaly detection, the framework enhances protection against cyber threats while aligning with Zero Trust Architecture principles. This proactive security solution serves as a critical asset for organizations striving to defend their virtual networks in cloud environments.

Keywords: Autoencoder, Cloud Security, Logistic Regression, Network Traffic Monitoring, Unauthorized Access.

Abstract: Hunger continues to be a global challenge, affecting millions of individuals despite advancements in food product and technology. This paper presents an AI-enabled food donation platform designed to grease real-time reporting of surplus food by donors. The system uses artificial intelligence to assay vacuity and match donations with the nearest NGOs and food banks grounded on position, demand, and urgency. Real-time notifications, efficient coordination, and analytics tools are integrated to ensure seamless operations and transparency. Food banks have the flexibility to accept or decline donations, allowing for optimized redistribution without assessing burdens related to storage or transportation.

Abstract: Accuplate is an advanced AI -controlled application that revolutionizes home cooking by autonomously creating adapted meals derived from the items present in the refrigerator or kitchen photos. The use of the sophisticated algorithm of the detection of Yolov5 objects, Accuplate accurately recognizes and categorizes a variety of food products, even in unfavorable environments, such as disorganized or poorly lit interiors of the refrigerator. After detecting folders, the system uses the Inception V2 model and natural language processing methods to provide recipe designs adapted to diet preferences and user limits. This smooth merger of computer vision and deep learning not only makes food preparation more efficient by providing fast and relevant recipe designs, but also supports sustainability by advising the use of available sources, reducing food waste. Accuplate improves ease and efficiency of daily cooking by removing the need for manual entry into ingredients or long recipe search. The solution works as an effective tool for kitchen automation and integrates advanced technology with user -focused design to support more intelligent and environmental cooking procedures. Index Terms: Object Detection, YOLOv8, FastAPI, TF-IDF, Word2Vec.

Abstract: This study presents a deep learning-based approach for automatic classification of scalp diseases and hair loss stages using image data. Leveraging convolutional neural networks (CNNs) with transfer learning, we evaluated multiple pre-trained models including ResNet50, VGG16, VGG19, and EfficientNet. Our method addresses challenges related to limited dataset size through image preprocessing and augmentation techniques, achieving high accuracy in distinguishing conditions like alopecia, psoriasis, and folliculitis, as well as hair loss progression stages. The trained models were integrated into a web application for user-friendly scalp condition diagnosis, enabling early detection and ongoing health monitoring.

Keywords: Deep Learning, Scalp disease, Hair Loss, Convolutional neural network (CNN).

Abstract: Sugarcane's overall productivity, yield, and crop health are all greatly impacted by nutrient deficiencies. Manual observation and laboratory testing are the mainstays of traditional methods for detecting these deficiencies, but they are costly, time-consuming, and frequently subject to human error. Furthermore, it can be difficult to make an accurate diagnosis because the visual symptoms of various nutrient deficiencies often overlap. This study suggests a deep learning-based method for automatically identifying nutrient deficiencies in sugarcane through image analysis in order to overcome these drawbacks. In order to accurately identify deficiencies like nitrogen, phosphorus, and potassium shortages, Convolutional Neural Networks (CNNs) are used to extract and classify features from images of sugarcane leaves. By offering scalable, precise, and real-time solutions, the suggested system improves efficiency by lowering reliance on laboratory testing and expert knowledge. By incorporating artificial intelligence into The goals of precision agriculture are to enhance crop management, maximise fertiliser use, and advance environmentally friendly farming methods. Results from experiments show how well deep learning models identify and categorise nutrient deficiencies, indicating their potential for practical agricultural uses.

Keywords: Disease detection, Image Processing, Deep Learning.

Abstract: The increasing prevalence of AI-generated deepfake images has become a significant concern in the context of misinformation and digital security. Deepfake technology, driven by generative adversarial networks (GANs) and sophisticated AI algorithms, enables highly realistic image alterations, making it challenging to differentiate genuine visuals from manipulated ones. This study introduces a deepfake image detection system utilizing convolutional neural networks (CNNs) for classification. By employing deep learning techniques, the system evaluates the authenticity of images and identifies alterations with high precision. Through training on diverse datasets, the model aims to bolster media integrity and strengthen digital security. The findings underscore the importance of reliable deepfake detection in minimizing the risks of manipulated content, offering valuable applications in fields such as journalism, social media verification, and digital forensics.

Keywords: Deepfake detection, image authenticity, machine learning, CNN, digital security, media verification, digital forensics

Abstract: The Blockchain-Based Lottery System delivers reliability and empirical services while ensuring transparency and fairness. Moreover, system security uses a tamper-only distributed counter and Smart Contracts. In addition, the Lottery System allows active participants to fully execute advanced bypass verification and authentication protocols which ensure system integrity. The participation rules of the lottery and the award distribution mechanisms are programmed in the smart contracts that govern the Ethereum blockchain which is the bedrock of the system. As a lottery system, it favors speed of transactions over delays when using Ethereum funds to disburse prizes and tickets. This system is flexible because of dependence on Ethereum’s robust ecosystem and scale easily, remaining the ideal fundamental for lottery frameworks that proliferate endlessly. It will focus on the user experience, the technology underpinning it, and the innovative horizon of the lottery systems based on blockchain technology to show how these systems can radically change the lottery ecosystem.

Keywords: Blockchain, lottery system, transparency, smart contracts, ethereum, distributed ledger, scalability, security, decentralized system, cryptocurrency.

Abstract: The increasing disparity in educational opportunities has sparked an urgent need for systems that can bridge the gap by ensuring that financial aid reaches the most deserving candidates. This paper introduces an AI-powered scholarship eligibility checker designed to automate and refine the process of scholarship evaluation. Leveraging state-of-the-art artificial intelligence techniques, machine learning algorithms, and robust data analytics, the system is engineered to identify qualified applicants accurately while mitigating human bias. By integrating comprehensive datasets including academic records, socioeconomic indicators, and historical scholarship data, this tool aims to not only expedite the evaluation process but also to enhance transparency in scholarship distribution. The survey explores the system’s architecture, the methodological framework, integration challenges, ethical considerations, and the potential impact on education equity.

Keywords: Artificial Intelligence, Scholarship Eligibility, Education Equity, Machine Learning, Data Analytics, Decision Support Systems.

Abstract: Agricultural producers, especially small-scale and marginal farmers, often face significant challenges related to unstable market access, price volatility, and exploitation by intermediaries. To address these systemic issues, this paper presents a comprehensive digital solution—Assured Contract Farming—which leverages technology to ensure transparent, enforceable agreements between farmers and buyers. The platform integrates legally binding digital contracts, secure payment mechanisms, real-time tracking, and reputation management to create a trustworthy ecosystem. By reducing uncertainties in procurement and enhancing contract enforcement, the system aims to empower farmers with reliable market access, fair pricing, and reduced dependency on middlemen. This approach aligns with the growing movement towards agricultural digitization and promises to contribute to sustainable rural livelihoods.

Keywords: Contract Farming, Digital Agreement, Market Access, Farmer Empowerment, Agricultural Technology, Supply Chain Transparency.

Abstract: This project presents the development of an Intrusion Detection System (IDS) using machine learning techniques to identify and classify potential threats in network traffic. Leveraging the NSL-KDD dataset, which provides a refined and widely accepted benchmark for network intrusion detection research, the system is trained to detect various types of attacks such as DoS, probe, R2L, and U2R. The project involves preprocessing the dataset, feature selection, and applying supervised learning algorithms like Decision Trees, Random Forest, and Support Vector Machines to build an accurate classification model. The goal is to enhance network security by enabling early detection of malicious activities and reducing false positive rates, ultimately providing a reliable and scalable solution for real-time threat detection in modern network environments.

Keywords: Intrusion Detection System (IDS), Machine Learning, NSL-KDD Dataset, Network Security, Supervised Learning, Random Forest, Feature Selection, Anomaly Detection, Cybersecurity, Attack Classification.

Abstract: The increasing globalization of education has led to a rising demand for scalable and efficient cloud-based translation solutions for academic materials. This survey paper explores the development of DocuLingo, an AI-powered document translation system leveraging AWS cloud services, particularly AWS Translate, to enhance accessibility in education. The study investigates the limitations of generic translation tools in handling domain-specific academic terminologies and proposes cloud-based customization strategies to improve translation accuracy. The primary objective is to evaluate how cloud-native AI translation can optimize academic content processing while maintaining cost-efficiency and scalability. Additionally, instead of building a standalone translation model, we assess methods to fine-tune and optimize existing cloud-native AI solutions for educational and technical documents. This survey highlights how cloud-native AI translation, specifically AWS Translate, can be optimized for academic use, ensuring higher accuracy and accessibility. By enhancing existing cloud-based AI models, we demonstrate how institutions can leverage scalable and cost-efficient translation solutions to break language barriers in research and education.

Keywords: Cloud Computing, AI-Powered Translation, Neural Machine Translation (NMT), AWS Translate, Academic Content Processing, Domain-Specific Translation, Language Accessibility, Cost-Efficiency, Scalability, Educational Technology., Employability.

Abstract: Coding interviews, hackathons held over the internet, and international software collaborations all necessitate platforms that provide real-time, interactive coding environments. But most of the available tools lack in providing smooth safe, and multi-language support for effective collaboration. In this paper, we introduce the Real-Time Code Collaborator (RTCC), a browser-based, full-stack system that allows multiple users to code, execute, and debug programs in real time collaboratively. RTCC integrates the strength of Web Sockets for live synchrony, Docker containers for secure running, and a React/Next.js frontend with voice and chat support. The platform not only increases productivity for remote teams but also enables greater accessibility in technical education and hiring. We cover the design principles, technologies, and architecture employed to build RTCC and contrast it with conventional tools such as Google Collab and VS Code Live Share.

Keywords: Real-time coding, Cloud IDE, Docker, Web Sockets, Collaborative development, Code execution,Multi-language support, OAuth 2.0, Git integration

Abstract: Dams are vital for water resource management, hydroelectric power generation, and flood control. However, traditional systems depend heavily on manual inspections and limited automation, which can delay critical responses to structural or environmental hazards. To address these challenges, this research proposes an Integrated Dam Automation System that combines IoT, image processing, and deep learning technologies for real-time monitoring and control. The system is built around an ESP32 microcontroller that gathers data from multiple sensors, including deep learning-based image analysis for crack detection, leakage sensors, pH and turbidity sensors for water quality assessment, and IoT-enabled sensors for water level monitoring and flood prediction. A dynamic, automated gate control mechanism is also included to regulate water levels effectively. With the addition of AI-powered predictive maintenance and remote monitoring through cloud integration, the system enhances the operational efficiency, responsiveness, and safety of dam infrastructure.

Keywords: Dam Automation, Deep Learning, Image Processing, Internet of Things, Real Time Monitoring.

Abstract: The internet offers a powerful medium for expressing opinions, emotions and ideas, using online platforms supported by smartphone usage and high internet penetration. Most internet posts are textual based and can include people’s emotional feelings for a particular moment or sentiment. Monitoring online sentiments or opinions is important for detecting any excessive emotions triggered by citizens which can lead to unintended consequences and threats to national security. Riots and civil war, for instance, must be addressed due to the risk of jeopardizing social stability and political security, which are crucial elements of national security. Mining opinions according to the national security domain is a relevant research topic that must be enhanced. Mechanisms and techniques that can mine opinions in the aspect of political security require significant improvements to obtain optimum results. Researchers have noted that there is a strong relationship between emotion, sentiment and political security threats. This study proposes a new theoretical framework for predicting political security threats using a hybrid technique: the combination of lexicon-based approach and machine learning in cyberspace. In the proposed framework, Decision Tree, Naive Bayes, and Support Vector Machine have been deployed as threat classifiers. To validate our proposed framework, an experimental analysis is accomplished. The performance of each technique used in the experiments is reported. In this study, our proposed framework reveals that the hybrid Lexicon-based approach with the Decision Tree classifier recorded the highest performance score for predicting political security threats. These findings offer valuable insight to ongoing research on opinion mining in predicting threats based on the political security domain.

Keywords: Political Security, Opinion Mining, Sentiment Analysis, Emotion Detection, Hybrid Lexicon-Based Approach, Machine Learning, Decision Tree Classifier, National Security.

Abstract: In the modern interdependent academic and professional landscape, finding the most appropriate domain expert is essential for cultivating collaboration, stimulating research, and industrial innovation. Yet, existing expert identification processes are typically disorganized, time-consuming, and decentralized. The Domain Expert Finder System (DEFS) rectifies these shortcomings by using web scraping to gather publicly accessible information from academic institutions, professional networks, and scholarly stores. This information is filtered, processed, and saved into a structured database, and users can access complete expert profiles through an interactive and searchable platform. In contrast to current systems that depend only on publication records or social networks, DEFS encompasses both faculty members and located students, hence expanding the range of expert discoveries. This paper overviews the system architecture, methodologies, comparative studies, and real-world implementations of DEFS. It talks about the major advantages like enhanced search accuracy, scalability of data, and automation along with ethical issues, technological hurdles, and prospects. Through a critical analysis of the literature and experimental verification, this survey points out the ways in which DEFS is an efficient and accessible solution to the issue of expert identification in contemporary knowledge environments.

Keywords: Expert Discovery, Web Scraping, Academic Profiles, Automation, Data Mining, Faculty Search, Placed Students, Knowledge Graphs

Abstract: Internet has made life of people so easy where people can access to any kind of information by just sitting at their homes. The foremost field that internet has covered is communication. When it is said communication based on internet, the first thing comes to everyone’s mind is E-mail. These are known to be most dependable way of communication. Voice feedback established virtual environment such as, screen readers help visually challenged individual gain access to internet applications tremendously. The benefaction made in project has licensed blind to forward and accept Email messages. This system can be handed-down productively by visually impaired and unlettered persons since it is based on TTS- Text to Speech, STT-Speech to Text conversions and IVR- Interactive Voice Response technologies.

Keywords: TTS-Text to Speech, STT-Speech to Text conversions and IVR-Interactive Voice Response.

Abstract: Managing hostel food services effectively is a complex task that involves meal planning, inventory control, and user satisfaction. Traditional manual methods often result in inefficiencies such as food wastage, inaccurate inventory tracking, and limited coordination between stakeholders. MealMap is a technology-driven solution that streamlines hostel food management by integrating automated meal adjustments, real-time inventory monitoring, and a structured feedback mechanism. Through role-based access for students, kitchen staff, and administrators, MealMap ensures a well-organized, data-driven approach to food service management. By leveraging real-time data analytics, MealMap optimizes resource utilization, minimizes waste, and enhances operational efficiency. Students can view meal plans and provide feedback, while administrators and kitchen staff can efficiently manage stock levels and adjust meal quantities as needed. The system’s structured approach improves accuracy, transparency, and sustainability, ultimately enhancing the overall dining experience in hostel environments. Future developments could incorporate AI-driven meal planning and mobile integration to further refine its impact on hostel food services.

Keywords: Meal Planning, Inventory Management, Automation, Feedback System, Sustainability

Abstract: The lack of structured interaction between alumni and students has long hindered the development of academic and professional networks in higher education. This paper surveys recent advancements in digital platforms designed to bridge this gap, with a particular focus on Alumni-Student Interaction Platforms (ASIPs). These systems aim to facilitate mentorship, career guidance, job sharing, and collaborative learning through integrated tools like online code compilers, discussion forums, and real-time messaging. The paper critically examines the architecture, features, and implementation challenges of such platforms, highlighting the importance of engagement, scalability, and data privacy. With technologies like React.js, Node.js, and Firebase becoming mainstream, these platforms now have the potential to become central hubs for academic communities.

Keywords: Alumni Interaction, Mentorship Platforms, Career Guidance, Web Development, Real-Time Communication, React.js, Node.js, Firebase, Networking Systems.

Abstract: Urban infrastructure maintenance is often hindered by inefficient complaint reporting systems, leading to delays in addressing critical public issues such as potholes, garbage accumulation, broken streetlights, and drainage problems. CivicFix is a cloud-based digital complaint system designed to simplify and automate the grievance redressal process. The platform allows users to report issues by uploading an image, while Google Maps API fetches the location details automatically. A machine learning model then classifies the complaint into categories such as potholes, garbage, streetlights, or drainage, ensuring that it is routed to the appropriate municipal department for resolution. The system features separate dashboards and logins for both users and department officers, allowing users to track complaint statuses and enabling authorities to efficiently manage and resolve issues. By leveraging cloud storage, AI-based classification, and automated routing, CivicFix enhances urban governance, making issue reporting more efficient, transparent, and community-driven. CivicFix not only simplifies complaint submission for users but also supports large-scale adoption by utilizing scalable technologies such as ReactJS, Firebase, and AI-powered classification. Its modular architecture allows integration with external APIs, real-time tracking services, and secure cloud storage. As a future-ready civic platform, CivicFix demonstrates how smart urban governance can be achieved through citizen-centric design, automation, and data-driven decision-making.

Keywords: Smart city, urban infrastructure, complaint redressal system, AI-based classification, Google Maps API, cloud computing, Firebase, Web-Based Application, Smart Governance

Abstract: Pneumonia is a potentially life-threatening respiratory infection that requires timely and accurate diagnosis for effective treatment. Traditional diagnostic methods, such as physical examination and radiologist interpretation of chest X-rays, are often time-consuming and susceptible to human error. This research presents an AI/ML-driven approach for automated pneumonia detection from chest X-ray images using advanced computer vision techniques. Leveraging convolutional neural networks (CNNs) and deep learning architectures, the system is trained and validated on a publicly available chest X-ray dataset. The model demonstrates high accuracy, sensitivity, and specificity in classifying pneumonia cases, thereby offering a reliable diagnostic aid. The integration of artificial intelligence in medical imaging not only accelerates the diagnostic process but also supports clinical decision-making, particularly in resource-constrained settings. This study highlights the potential of AI-powered tools in enhancing diagnostic efficiency and contributing to the broader goal of intelligent healthcare systems.

Keywords: Pneumonia Detection, Convolutional Neural Networks (CNN), Deep Learning, Medical Imaging, Image Classification, DICOM Processing, Healthcare Diagnostics, Lung Infection Detection, Computer Vision, Django Framework, Radiology Support System, Bounding Box Localization, AI-Driven Medical Analysis.

Abstract: This research paper considers the development of medical online bookings using intelligent chatbots to improve the user experience and simplify the health process. The platform allows users to plan effective medical appointments, order medications, and receive medical information. Users with AI and chatbots will receive immediate support, personalized recommendations and notifications for a smoother and more user-friendly experience. This study examines chatbot technology integration, impact on user interaction, and potential platforms to improve delivery and performance of healthcare systems, schedules, and living procedures.

Keywords: Online medical appointments, health technology, Planning, Online Pharmacy, Pet Medicine, Prescription Verification.

Abstract: This research proposes a comprehensive system designed to detect stress levels among IT professionals by leveraging real-time facial analysis powered by advanced machine learning techniques. The underlying concept is based on the psychological understanding that human emotions are visually expressed through subtle facial movements and micro-expressions, making them valuable indicators for assessing an individual’s mental and emotional state. By utilizing these non-intrusive cues, the system offers a privacy-conscious and continuous approach to psychological monitoring without requiring active user participation. At the core of the system is the use of Convolutional Neural Networks (CNNs), which are highly effective in processing and interpreting visual data, particularly for emotion recognition tasks. The Deep Face library is employed to extract deep feature representations from facial images, enabling accurate classification of emotions that correlate with varying levels of psychological stress. For initial face detection and localization, classical Haar Cascade classifiers are integrated, providing reliable identification of facial regions within both static images and live video streams. The implementation includes a web-based interface developed using the Django framework, which allows users to interact with the system in real time. This interface supports continuous webcam input, emotion-based feedback display, and optional logging of stress assessments, ensuring a user-friendly experience suitable for deployment in organizational settings. Experimental evaluations were conducted using both publicly available emotion datasets and live webcam feeds to validate the system's effectiveness. The results indicate high accuracy and consistent performance in classifying emotional states and estimating corresponding stress levels. These findings underscore the system's potential as a practical tool for real-time mental health monitoring in professional environments, particularly within the high-pressure context of the IT industry.

Keywords: Stress Identification, Facial Emotion Analysis, Machine Learning, Deep Face, CNN, Real Time Monitoring, IT Workforce.

Abstract: A significant backlog of unresolved cases currently plagues the Indian legal system, mostly as a result of ineffective case management procedures and a lack of established procedures for setting case priorities. The scheduling and prioritization procedure is now handled manually by court employees, who rely on their own experience and subjective judgment. In addition to wasting important time and money, this manual intervention causes delays and dealing with inconsistencies urgent issues. when an automated solution is desperately needed to expedite the prioritization process and guarantee that important cases receive prompt attention while the number of pending cases keeps growing. The creation of a software-based tool that automates the process of prioritizing and ranking court cases is suggested in this paper: the Judicial Case Priority Management System. To give each case a priority ranking, the algorithm makes use of key legal characteristics as well as portions of the Indian Penal Code (IPC). Case information, IPC sections, and the accompanying priority scores are kept in a special database. The algorithm at the heart of the system determines each case's severity and urgency, resulting in a final priority score on a scale from 0 to 10. High-profile or urgent instances that require immediate care are indicated by a score of 0, whereas less urgent cases are indicated by a score of 10. To determine the priority score, the algorithm takes into account a number of variables, including the type of violation, related IPC sections, prior case history, and other contextual information. In order to ensure that courts can handle high-priority cases quickly, the calculated scores are saved in the database and utilized to dynamically categorize cases. By minimizing human participation and increasing speed, this structured prioritizing lowers the possibility of bias and error in the case scheduling process.

Keywords: Judicial case management, Priority Scoring, Legal Automation, Court Efficiency, Indian Penal Code.

Abstract: This thesis discusses the convergence of banking services and the changing needs of platform businesses— variable, data-intensive models with special financial needs. Conventional banks, bound by legacy systems and stiff compliance, are unable to meet these needs. By means of qualitative research, such as expert interviews and secondary data, the report finds the most important service gaps and how fin techs are filling them with API-based platforms, AI compliance, and blockchain for scalable, efficient solutions. The study utilizes Disruption Theory and Platform Business Theory to examine opportunities and threats facing banks, providing hands-on advice like adopting fintech collaborations, embedded finance, and tech upgrades to remain competitive in a fast-evolving digital environment.

Keywords: Platform businesses, fintech, embedded finance, API-based solutions, real-time payments, digital transformation, compliance, disruption theory.

Abstract: In this study, an AI-powered fashion recommendation system that integrates image processing and real-time weather data to offer tailored outfit recommendations is investigated. The system allows users to upload images of clothing, which are processed using convolutional neural networks (CNN) for feature extraction such as color, texture, and garment type. At the same time, weather data is accessed through the OpenWeatherMap API. A rule-based and optionally machine learning-augmented recommendation engine thereafter recommends weather-matched ensemble combinations. The solution improves the decision-making of the user by considering personal style and environmental factors, offering context-sensitive and smart wardrobe guidance. The system does not only aim at fashion-conscious users but also at fashion retailers and online stores looking to boost customer interaction through intelligent clothing recommendations based on local weather patterns and individual tastes. Integration of AI and image processing in fashion bolsters the user experience through automation of choosing an outfit and making sure that the garments match both functional and aesthetic requirements. The model supports sustainable fashion in terms of better use of current wardrobes and minimized unwanted purchases. The study also assesses the scalability and flexibility of the system for various demographics. of users and worldwide meteorological conditions, thereby creating a framework for future developments in fashion technology.

Abstract: Medical datasets frequently contain missing values, which can negatively impact machine learning models used in healthcare. However, imputing these values while ensuring patient privacy presents a significant challenge. This survey explores various privacy-preserving data imputation techniques, with a focus on Secure Multi-Party Computation (MPC). We review four imputation methods—mean, median, regression, and k-nearest neighbors (KNN)—and how each can be implemented securely in distributed medical environments. The paper also discusses hybrid approaches, integration with differential privacy, and federated settings. Our analysis concludes that MPC-based imputation provides strong privacy guarantees with high accuracy, paving the way for privacy-conscious medical data analysis.

Keywords: Data Imputation, Medical Data Privacy, Multi-Party Computation (MPC), Secure Computation, Privacy-Preserving Machine Learning

Abstract: The exam Hall Seating Arrangement System is developed for the college to simplify examination hall allotment and seating arrangement. It facilitates access to the examination information of a particular student in a particular class. The purpose of developing an exam hall seating arrangement system is to computerize the traditional way of conducting exams. Another purpose for developing this software is to generate the seating arrangement report automatically during exams at the end of the session or in between sessions.

Abstract: Log analysis is the systematic process of collecting, interpreting, and analysing log data generated by various systems, applications, devices, and networks. Logs are automatically produced records that document system events, user actions, errors, performance metrics, security incidents, and other activities critical to the functioning of IT environments. Through log analysis, organizations can gain valuable insights that enhance operational efficiency, bolster security, and ensure regulatory compliance. The core goal of log analysis is to convert raw log data into actionable information that helps in troubleshooting issues, identifying performance bottlenecks, detecting security threats, and optimizing system resources. One of the primary motivations for log analysis is its utility in troubleshooting and diagnostics. Logs capture comprehensive details about system events, errors, crashes, and service disruptions, which are essential for identifying the root causes of issues. By analysing logs, IT administrators can gain a better understanding of how systems behave under normal and abnormal conditions. This enables them to pinpoint the exact causes of failures or performance degradations, allowing for timely resolution of problems and reducing system downtime. Furthermore, log analysis supports proactive monitoring by enabling real-time detection of anomalies, such as unusual spikes in resource usage, errors, or service response times. This helps organizations identify potential problems before they impact end user.

Keywords: Log Analysis, Machine Learning, NLP, Random Forest Algorithm, Root Cause Analysis, Visualization, Log Parsing.

Abstract: Blockchain technology holds the potential to transform the agricultural and food supply chain by introducing transparency and automation through smart contracts and other integral blockchain features. This paper explores the functionality of blockchain systems, their possible integration into existing Supply Chain Management (SCM) frameworks, and the implications for legal and regulatory bodies. The increasing adoption of blockchain challenges traditional institutions, especially government entities historically trusted with transaction verification. Therefore, the Agri-Food sector demands a robust, transparent, and reliable system that ensures traceability and efficient product flow. This study aims to highlight how blockchain can bring a paradigm shift to existing systems.

Keywords: Agricultural product, food delivery, consumer, NGO, web application

Abstract: Computer Vision Syndrome (CVS) is a growing public health concern caused by the increased use of digital defenses in everyday life, particularly among working professionals and scholars. Symptoms similar as eye strain, blankness, blurred vision, and headaches are generally reported due to extended screen exposure. Addressing these challenges, this study introduces EYELUME, an innovative,non-invasive system that leverages the VIT- Pupil model a Vision Motor (VIT)- grounded armature for the accurate segmentation and analysis of pupil images. The VIT- Pupil model is specifically designed to handle noisy and low- resolution images, making it largely suitable for real- world operations where ideal imaging conditions can’t always be guaranteed. Unlike traditional Convolutional Neural Network (CNN) approaches, which struggle to capture global dependences in visual data, VIT models exceed in landing contextual and spatial information throughout the image. By tracking pupil variations over time, EYELUME enables real- time monitoring of digital eye strain symptoms. The model achieves an emotional segmentation delicacy of 99.6, thereby offering a dependable foundation for early discovery of CVS and enhancing digital eye health monitoring systems.

Keywords: Vision Mills, Computer Vision Syndrome, VIT- Pupil, Pupil Segmentation, Deep Learning, Pupillometry, Digital Eye Health.

Abstract: This paper presents an integrated health informatics framework aimed at forecasting and mitigating dengue outbreaks by fusing hydro-climatic conditions with symptom-based user inputs. The proposed system evaluates environmental variables—namely rainfall, temperature, and humidity—alongside self-reported symptoms and historical disease trends to estimate the likelihood of dengue transmission across different regions. By implementing rule-based logic and leveraging spatio-temporal analysis, the system offers real-time risk estimations and context-specific preventive guidance. Key features include a symptom evaluation tool, geo-based alert notifications, and a risk scoring mechanism derived from user questionnaires. The interface also incorporates visual elements such as dynamic heatmaps and time-series charts to enhance public understanding and actionable response. Developed with usability and public health relevance in mind, this platform serves as a practical and scalable tool to support early intervention strategies against vector-borne diseases like dengue

Abstract: Intracranial Hemorrhage (ICH) is a serious and potentially fatal condition marked by bleeding in the cranial cavity. It often arises from trauma, high blood pressure, or vascular issues. Early detection of ICH is crucial for improving patient outcomes and lowering mortality rates. Computed Tomography (CT) is the standard method for quickly diagnosing ICH due to its widespread availability and high sensitivity to acute bleeding. Despite this, radiologists must manually interpret CT images, which is labor-intensive and time-consuming, leading to variations in accuracy. Recently, deep learning, especially Convolutional Neural Networks (CNNs), has become a valuable tool for automating medical image analysis. This study looks at how deep learning techniques can automatically detect and classify ICH in brain CT scans. We review existing models, discuss data preprocessing methods, evaluate performance metrics, and highlight commonly used datasets like RSNA and CQ500. We also tackle challenges such as data imbalance, model interpretability, and clinical integration. Our findings show that deep learning models can achieve high diagnostic accuracy and significantly enhance clinical decision-making in emergency situations. Future research should aim to improve model generalization, explainability, and real-time deployment in clinical settings.

Keywords: Intracranial Hemorrhage, CT Scan, Deep Learning, Convolutional Neural Networks, Medical Image Analysis, ICH Detection, Automated Diagnosis, RSNA Dataset, CNN, Healthcare AI

Abstract: One of the most common gynaecological cancers affecting women globally is endometrial cancer, which develops from the lining of the uterus. The prognosis is improved by early diagnosis, but traditional techniques like biopsy and ultrasound are frequently intrusive, costly, or inaccessible. Recent developments in artificial intelligence, specifically in the areas of machine learning (ML) and image processing, present promising instruments for the automated, non-invasive, and precise detection and grading of endometrial cancer. This survey investigates how to improve the quality of histopathological images for analysis using image preprocessing methods like RGB to grayscale conversion, noise reduction, thresholding, segmentation, and feature extraction. Additionally, it assesses how well deep learning models—particularly Convolutional Neural Networks (CNNs) and transfer learning techniques—classify malignant tissues and forecast tumour stage. and

Keywords: Endometrial Cancer, Image Preprocessing, Machine Learning, CNN, Histopathology, Medical Imaging, Transfer Learning, Deep Learning.

Abstract: This paper introduces the Med-Crop Recommendation system, a smart farming advisory platform aimed at supporting farmers in South Karnataka in cultivating medicinal crops optimally. The platform leverages data analytics and machine learning techniques to provide personalized crop recommendations based on inputs such as soil health, climate conditions, water availability, and geographical data. In a region marked by diverse agro-climatic zones and growing market interest in herbal products, this system promotes sustainable farming by encouraging the adoption of low-water, high-value crops. The platform is designed with accessibility and scalability in mind, targeting small to medium-scale farmers.

Keywords: Smart Farming, Medicinal Crops, AI in Agriculture, Soil Analysis, Climate Data, Crop Recommendation, Sustainable Agriculture, Precision Farming.

Abstract: In today’s technologically advancing academic environment, the timely and accurate evaluation of subjective answers plays a crucial role in educational assessment. While objective-type answers can be easily evaluated using automated systems, the assessment of subjective responses demands more sophisticated techniques that consider context, content relevance, structure, and grammatical accuracy. This research paper presents an AI- based assessment framework capable of evaluating both handwritten and typed subjective answers using Machine Learning (ML) and Natural Language Processing (NLP) methods. Handwritten responses are first digitized using Optical Character Recognition (OCR), which converts the input into textual data. Subsequently, the evaluation process utilizes semantic similarity measures, keyword extraction, and grammatical analysis. The framework integrates pretrained language models along with custom-trained classifiers to compare student responses against reference solutions, enabling the assessment of contextual accuracy and logical coherence. The proposed method reduces human bias, enhances consistency through algorithmic scoring, and significantly minimizes manual grading effort. The results this work demonstrates that the system achieves high accuracy (99.30%) and outperforms traditional evaluation techniques in speed and reliability. This framework offers a scalable and intelligent approach for subjective answer assessment, contributing meaningfully to the integration of AI in modern education.

Keywords: Optical Character Recognition (OCR), Convolutional Neural Networks (CNN), Machine learning(ML) Natural Language Processing (NLP), Large Language Models (LLM), Subjective Answer Assessment.

Abstract: Coal mines are prone to dangerous conditions such as gas leaks, structural failures, water flooding, and lack of proper communication during emergencies. This paper presents "Terra Rover," a compact, semi-autonomous six-wheel rover system built specifically for underground coal mine safety and rescue operations. It is designed to detect hazardous environments using an array of sensors (gas, water, temperature, ultrasonic crack detection), mark unsafe zones, and transmit live data to rescue teams. The rover is capable of navigating rough terrains with a custom mechanical suspension and can deliver rescue kits while maintaining real-time communication. Our system uses Arduino Mega and Esp32, integrated with wireless transmission and cloud logging. The rover also includes a fallback system to transmit last-known coordinates in case of destruction. The result is a robust, field-deployable, affordable platform for early hazard detection and safety support in coal mining zone

Keywords: Coal Mine Safety, Exploration Rover, Gas Detection, Rescue Robot, Hazard Mapping, Underground Communication, Environmental Monitoring, Disaster Response Robot, Remote Sensing.

Keywords: Natural Language Processing, Legal AI, Transformer Models, BERT, Indian Laws, Legal Advisor.

Abstract: Face recognition is widely used in security, authentication, and surveillance. However, recognizing partially occluded faces remains a significant challenge due to missing facial features. This project proposes a 3D Partially Occluded Face Recognition System Using Hybrid Deep Learning Techniques, integrating 3D geometric facial structure, texture analysis, and advanced deep learning models to improve recognition accuracy in occluded scenarios. The system employs ResNet50 for robust 2D feature extraction, while PointNet++ processes 3D facial point cloud data. To mitigate the impact of occlusions such as masks, sunglasses, and scarves, selfattention mechanisms and transformer-based CNNs are used to focus on unoccluded facial regions. Additionally, feature-level fusion combines 3D structural features with facial texture to enhance performance. A diverse dataset, including BU-3DFE, FRGC v2.0, Bosphorus, and FaceWarehouse, is used for training and evaluation. The system is tested across various occlusion types to ensure robustness, reliability, and high recognition accuracy. Performance metrics such as recognition accuracy, F1 score, and occlusion robustness score are used for evaluation. For real-world deployment, the system is integrated into a web-based application the proposed system significantly improves face recognition accuracy under occlusions, making it a practical solution for security and authentication applications

Keywords: 3D Face Recognition, Occlusion Handling, Deep Learning, Feature Fusion, Hybrid Recognition Techniques, Web Deployment.

Abstract: Parkinson’s Disease (PD) is a progressive neurological disorder that affects movement and coordination, often diagnosed at later stages due to subtle early symptoms. Early and accurate detection of Parkinson’s Disease is crucial for timely intervention and improved quality of life. This project explores the application of machine learning techniques to detect Parkinson’s Disease using biomedical voice measurements and other relevant features. By training classifiers such as Support Vector Machines (SVM), Random Forest, and K-Nearest Neighbors (KNN) on datasets containing patient voice data and clinical attributes, the system learns to distinguish between healthy individuals and those with PD. Feature selection and data preprocessing are employed to enhance the model's accuracy and reduce overfitting. The results demonstrate that machine learning models can effectively support medical professionals in diagnosing Parkinson’s Disease, offering a non-invasive, cost-effective, and automated approach to early detection. This study highlights the potential of artificial intelligence in transforming traditional diagnostic processes in neurology.

Abstract: This research focuses on developing a robust video stabilization technique to minimize jittery motion in video footage. The proposed method employs optical flow, utilizing the Lucas-Kanade algorithm, to estimate motion between consecutive frames. Affine transformations are applied to align the frames by calculating geometric corrections, ensuring smoother transitions. To further enhance stability, trajectory smoothing is incorporated, which refines the motion corrections and reduces abrupt changes. The study also explores the mathematical principles behind the key processes, including motion estimation and geometric transformations. Furthermore, strategies to optimize the method for high-resolution videos are discussed, emphasizing both computational efficiency and visual enhancement. Experimental evaluation confirms that the proposed approach effectively stabilizes video sequences, making it a practical solution for handheld or dynamic video applications.

Keywords: Video stabilization, Optical Flow, Lucas-Kanade, Affine transformation, Trajectory smoothing, Motion estimation, High-resolution optimization.

Abstract: Access to essential services is a vital need for students who migrate to other places. While access to essential services is crucial for students relocating, simply providing access falls short of fostering a sense of belonging. Existing platforms lack localized information as well as personalization. This paper proposes an open platform that streamlines access to campus services while offering a personalized experience through machine learning. The platform streamlines access to services such as housing, dining, and marketplace while providing a personalized experience. The platform provides an interface for both students and service providers, allowing providers to gain insights and improve their businesses as well. The proposed system tackles common drawbacks faced by existing recommendation systems by employing a hybrid recommender system. The system follows a service-oriented architecture developed using microservice architecture, which allows services to be independent and makes the platform scalable. The platform addresses limitations of traditional recommendation systems by utilizing a hybrid approach, which results in better accuracy in recommendations than existing systems.

Keywords: Campus Services, Machine Learning, Recommendation systems, hybrid recommendation, personalization, open platform, scalability, service-oriented architecture.

Abstract: Cutaneous diseases rank as a leading global health issue and many of them should be diagnosed in time to treat them appropriately. With the development of deep learning, automated skin disease diagnosis is now possible and has been improved to be more accurate. In this paper, we propose a deep-learning methodology based on the VGG16 CNN model for classifying skin diseases from the DERMNET dataset. Preprocessing and data augmentation steps are employed to enhance the robustness and generalization ability of the model. The above system effectively demonstrated a diagnostic accuracy of around 90% indicating that it can provide great support for dermatologists and reduce diagnostic errors. In addition, to provide real-time diagnostic assistance, a Streamlit tool is implemented as an interface to invoke the trained model.

Keywords: Deep Learning, Skin Disease Detection, VGG16, Convolutional Neural Network, Data Augmentation, Stream lit.

Abstract: Intrusion Detection Systems (IDS) are critical for mitigating evolving cybersecurity threats. This study investigates the integration of Machine Learning (ML) and Deep Learning (DL) techniques to enhance IDS efficiency. A dual-panel IDS is developed, incorporating an attack detection module for user uploads and an admin panel for model training and testing. The system leverages multiple classification algorithms, including Support Vector Machine (SVM), Random Forest, XGBoost, AdaBoost, and Decision Tree, to improve intrusion detection accuracy. A dynamic model selection mechanism is implemented to optimize algorithm performance at runtime, complemented by graphical visualizations for comprehensive threat analysis. Various IDS datasets are evaluated to assess detection effectiveness, addressing challenges such as computational complexity and real-time traffic management. Experimental results indicate an accuracy range of 92% to 96%, with Random Forest and Decision Tree performing optimally based on dataset characteristics. This research contributes to the advancement of IDS by improving detection reliability, reducing false positives, and enhancing system scalability, ultimately strengthening cybersecurity defenses.

Keywords: IDS, ML, DL, Network Security, Random Forest, SVM, Cybersecurity, Anomaly Detection, False Positives, Scalability, Accuracy, XGBoost, Decision Tree.

Abstract: In an era of increasing digital connectivity, the sophistication and frequency of cyberattacks have grown exponentially, rendering traditional rule-based intrusion detection systems (IDS) insufficient. This literature survey explores the recent advancements in AI-powered IDS solutions, with a particular focus on machine learning (ML)-driven approaches for harmful packet detection. The review analyzes 25 recent research papers published between 2020 and 2025, highlighting trends in model development, dataset utilization, real-time deployment, edge computing, and automation in threat response. While many existing systems achieve high detection accuracy using algorithms such as Random Forest, SVM, CNN, and ensemble techniques, they often fall short in critical areas—such as real-time performance, attack simulation, automated remediation, and handling minority class attacks. This survey identifies those gaps and establishes the motivation for a lightweight, modular IDS that not only detects but also responds to intrusions through intelligent patch recommendations. By comparing existing approaches and their limitations, the paper lays the foundation for building adaptive, scalable, and semi-autonomous security solutions suitable for modern network environments.

Keywords: Intrusion Detection System, Machine Learning, NSL-KDD, Network Security, Automated Patching, Real-Time Threat Detection, Cyberattack Classification, Lightweight IDS

Abstract: Osteoporosis is a progressive bone disorder that reduces bone density and deteriorates bone structure, significantly increasing fracture risk. Traditional diagnostic tools like DEXA scans are often expensive and inaccessible, particularly in rural or under-resourced regions. This project introduces an AI-based diagnostic system designed to detect osteoporosis using X-ray images and clinical data. The system features three main modules: a CNN model trained to classify spine and knee X-rays, a machine learning-based clinical predictor using patient data, and a stage detection module for assessing disease severity. The CNN model achieved approximately 95% accuracy, while the clinical predictor using Gradient Boosting reached 92.01%. A Flask-based web application provides an easy-to-use interface for patients and healthcare professionals. The system also delivers personalized treatment recommendations and optional doctor consultation links. By combining image analysis and clinical data evaluation, this hybrid approach offers a cost-effective, accessible, and accurate tool for early osteoporosis detection, especially beneficial in underserved areas.

Abstract: Seed quality assessment is a fundamental step in ensuring high agricultural productivity and food security. However, traditional methods for evaluating seed quality—relying on manual inspection and mechanical processes—are time-consuming, labor-intensive, inconsistent, and often prone to human error. In this project, we introduce a deep learning-based approach that utilizes Convolutional Neural Networks (CNNs) in combination with the YOLOv5 object detection algorithm to automate and enhance the seed quality grading process. The proposed system focuses on five commonly used food pulses: maize, rice, beans, channa, and wheat. By analyzing characteristics such as size, shape, texture, and color from high-resolution images, the model identifies and classifies seeds into three distinct categories: Grade A (good), Grade B (fair), and Grade C (poor). The implementation leverages Python, PyTorch, Flask, and OpenCV for data preprocessing, model training, interface development, and live camera-based inference. Real-time performance is achieved using a lightweight Flask-based GUI that enables users to conduct seed analysis via webcam with instant feedback. The model demonstrates high reliability and accuracy—achieving a performance score of 92%—even under varying lighting conditions and image quality. The system is optimized to run on low-resource devices, making it deployable in field environments as well as small-scale processing units. This intelligent solution addresses a critical need in precision agriculture by significantly reducing human effort, improving consistency, and increasing the speed and efficiency of seed sorting. It serves as a scalable, low-cost, and practical tool that can be extended to other crop varieties, contributing toward the modernization and automation of agricultural practices.

Keywords: Seed Quality, CNN, YOLOv5, Deep Learning, Image Processing, Agriculture AI

Abstract: The increasing sophistication of urban roadways necessitates innovative approaches to tackle traffic congestion, enhance public safety, and streamline emergency interventions. This project introduces a YOLOv8-powered intelligent traffic system specifically engineered to identify emergency vehicles (like ambulances and fire trucks) for priority lane access. Crucially, it also features accident detection and immediate alert capabilities.

Keywords: Research Paper, Technical Writing, Science, Engineering and Technology

Abstract: Agriculture near forests faces threats from wild animals and fire, causing crop loss and financial strain. Traditional methods like manual monitoring are often ineffective. The "Innovative Solutions for Wildlife Conflicts Mitigation" offers an automated solution using sensors, fire detectors, AI-based animal recognition, and real-time alerts via SMS, LEDs, and alarms. It integrates Raspberry Pi, Arduino Nano, cameras, and deterrents for efficient threat detection. AI analyzes camera feeds to classify animals and trigger targeted responses. Field tests show improved safety, reduced damage, and a cost-effective, scalable approach to protect farms sustainably.

Keywords: Wild animal detection, Fire alert system, AI-based monitoring, Real-time alerts, Agricultural protection, Raspberry Pi integration.

Abstract: This study explores thyroid disease detection using K-Means and Fuzzy C-Means clustering algorithms. By analyzing patient data, the models classify thyroid conditions efficiently. Comparative evaluation highlights accuracy and effectiveness, aiding early diagnosis. The research emphasizes the importance of machine learning in medical diagnostics, enhancing predictive capabilities for thyroid disorders. The data is first pre- processed, selected, extracted and then classified defect or normal class. This algorithm gives best result for overlapped data also. Existing studies focus on the binary classification tasks. This study improves on prior experience with conventional CNN models (i.e., VGG models), the more advanced CNN architecture, the Exception model was implemented and compared to achieve the automatic diagnosis with increased efficiency and accuracy. It is also likely that the established structure can be easily translated to determine the diagnosis of other disease

Keywords: Thyroid Disease, Data Mining, Fuzzy C Means Clustering.

Abstract: Kinesiology, the scientific study of human movement, integrates anatomy, physiology, biomechanics, and motor learning to understand the principles of movement. Fundamental movements—such as walking, running, jumping, lifting, pushing, pulling, and throwing—form the basis of more complex motor skills and physical activities. A kinesiological analysis examines these movements to identify joint actions, muscular involvement, planes of motion, axes of rotation, and the forces acting upon the body. This paper presents a comprehensive kinesiological analysis of fundamental human movements. Each movement is deconstructed by evaluating the biomechanical and anatomical components, focusing on joint kinematics, muscle activation, and force generation. For instance, during walking, coordinated activity between the hip flexors, knee extensors, and ankle plantarflexors allows forward propulsion while maintaining balance. In a vertical jump, triple extension at the hip, knee, and ankle, powered by concentric muscle contractions, is critical for maximizing lift-off force. The analysis of throwing reveals coordinated sequential activation, from the lower limbs to the upper extremities, emphasizing the kinetic chain's role. Understanding the kinesiological basis of these movements has practical implications in physical education, sports science, rehabilitation, and ergonomics. It allows professionals to enhance performance, prevent injuries, and design movement interventions tailored to individual needs. Through integrated kinesiological knowledge, fundamental movements can be optimized for various population groups, including athletes, patients, and sedentary individuals, fostering efficient and safe movement practices.

Keywords: Kinesiology, Fundamental Movement, Joint Kinematics, Muscle Activation, Biomechanics, Human Motion

Abstract: The project shows a way to use Machine Learning (ML) to find Autism Spectrum Disorder (ASD) early on, acknowledging the challenges of diagnosing the condition while striving to mitigate its severity through early interventions. The suggested system uses four typical ASD datasets, ranging from infants to adults, to test four Feature Scaling (FS) techniques: Quantile Transformer, Power Transformer, Normalizer, and Max Abs Scaler. Included scaled datasets are used for machine learning computations (like K-Nearest Neighbors, Gaussian Naïve Bayes, Logistic Regression, SVM, LDA, Ada Boost, and Random Forest). Factual estimations used to Find the best FS methods and classifiers for each age group. Babies, children, adolescents, and adults are the groups for which the voting classifier most accurately predicts ASD. The assignment includes an analysis of the relevance of a specific aspect. Employing four Component Determination Strategies to help medical care professionals with ASD screening and to emphasize the importance of calibrating machine learning approaches in predicting ASD across age groups. The suggested structure outperforms the existing early ASD finding methods. A group process that used a Voting Classifier with Random Forest (RF) and AdaBoost was able to get 100% accuracy, which made ASD recognition even stronger and more accurate.

Keywords: Machine Learning, Classification, Autism Spectrum Disorder, Feature Scaling, and Feature Selection Methods.

Abstract: In today's digital age, cybersecurity incidents are rising in frequency and sophistication, requiring intelligent, responsive, and proactive solutions. This paper presents the implementation of an AI-Driven Cybersecurity Chatbot for Incident Response, integrating OpenAI for natural language understanding and VirusTotal for real-time threat intelligence. The system assists users in detecting, analyzing, and responding to threats through a user-friendly conversational interface.A core feature is the secure user authentication system with login and logout tracking, where activity is logged and monitored across devices. If a login is detected from an unknown or suspicious device, the system sends an instant alert to the original device or email account, helping to prevent account takeovers. Furthermore, it provides actionable insights on how the email or device may have been targeted, such as brute-force attempts or phishing-based intrusions, enhancing transparency and user awareness.The system also includes MongoDB-backed chat history, a modern UI, and structured responses with contextual definitions and safety advice for AI and cybersecurity terms. By combining real-time intelligence with AI-powered guidance and proactive alerting, the chatbot empowers users to detect, understand, and mitigate cyber threats effectively.

Keywords: Artificial Intelligence (AI), Cybersecurity, Incident Response,Natural Language Processing (NLP), Chatbot, VirusTotal API, Threat Intelligence, OpenAI Integration,Login/Logout Tracking,Cross-Device Alert System

Abstract: Tuberculosis (TB) has remained a significant global health concern for a very long time. It necessitates accurate and early detection and curing to improve patient health. This study comprehensively reviews various methods employed for detecting TB using chest X-rays. It explores traditional diagnostic approaches, including manual interpretation by radiologists, and advances in automated techniques such as machine learning (ML) and deep learning (DL) algorithms. The paper highlights the strengths and limitations of different methodologies, focusing on their accuracy, sensitivity, specificity, and computational efficiency. This study aims to offer insights into the revolution of TB detection methods and inform about developing more robust and scalable tools.

Keywords: Tuberculosis, Deep learning, Machine learning, Early detection.

Abstract: The increasing number of vehicles on roads, coupled with urban population growth, has significantly intensified traffic-related challenges, including congestion, delays, and accidents. Traditional traffic management systems often lack the adaptability and responsiveness required to ensure road safety in real time. This report explores the design and implementation of an Intelligent Traffic Safety System (ITSS) that leverages modern technologies such as Yolo model-v3, Artificial Intelligence (AI) and computer vision to enhance traffic monitoring, management, and accident prevention.

Keywords: Research Paper, Technical Writing, Science, Engineering and Technology

Abstract: Cloud-based Enterprise Resource Planning (ERP) systems promise simplified operations and increased efficiency. However, their acceptance in Kenyan universities is still in its initial stage of implementation due to financial limitations, legacy system incompatibilities, and a lack of specialized expertise. Additionally, integrating ERPs with cloud services presents complex security challenges. The study highlighted widespread concerns regarding data breaches, unauthorized access, and compliance with Kenyan data protection laws, as well as the Data Act of 2019. The study found a substantial security-usability trade-off, where the need to maintain strong data protection conflicts with the demand for increased functionality through cloud integration. This study proposed the Adaptive Trust Model in response to these difficulties and with direct input from the empirical results. This paradigm offers Kenyan universities a safe and contextually appropriate framework for combining cloud services and ERPs. The development of this model is crucial, as the findings indicated the inadequacy of generic security solutions in addressing the unique infrastructural and regulatory landscape of Kenyan universities. A comprehensive literature review was conducted to analyze past contributions and contextualize the study's findings. Cluster sampling was employed to select a representative sample of universities, and data analysis was performed using R software, facilitating both quantitative and qualitative insights.

Keywords: ERP, cloud computing, secure integration, Kenyan universities, data protection, Adaptive Trust Model, digital transformation.

Abstract: One of the biggest challenges facing blind assistance systems is how they can navigate with safety and independence in such complicated real-world scenarios, given that traditional tools that assist these users are usually simplistic. Among many techniques that emerge as essential to upgrading these systems are machine learning and deep learning. These methods introduce considerable object detection, voice recognition, and distance measurement capabilities. This review summarizes the findings of recent studies in the application of neural networks, such as convolutional neural networks (CNNs), and advanced models in real-time object recognition and environmental awareness. Models like Faster R-CNN, SSD, and DenseNet have shown exceptional performance in object detection and segmentation with high accuracy rates and reliability. However, the challenges include diversity in datasets, limitations in real-time processing, and user adaptability. Furthermore, computational efficiency and optimizing deep learning models for low-power devices remain crucial areas for improvement. Enhancing multimodal feedback, integrating adaptive learning models, and improving response time are essential for real-world deployment. This review represents a great step forward in assistive technology, providing real-time, reliable feedback to help visually impaired users navigate their surroundings with greater independence and confidence.

Keywords: Visually Impaired, Computer Vision, Deep Learning, Object Detection, YOLO Algorithm, Real time.

Abstract: Digital marketing has evolved significantly with the integration of artificial intelligence and machine learning technologies, yet existing approaches often treat Search Engine Optimization (SEO), Social Media Optimization (SMO) and digital marketing as separate entities. This research presents a novel meta-level programming framework that intelligently integrates these components through automated optimization algorithms. The proposed framework leverages machine learning techniques to dynamically adjust marketing strategies based on real-time performance metrics and user behavior patterns. We conducted comprehensive experiments using the UCI Online Retail II dataset containing 1,067,371 transactions and the Marketing Campaign Performance dataset with 200,000 campaign records. Our meta-programming approach demonstrated a 34.2% improvement in conversion rates compared to traditional methods with SEO performance increasing by 28.7% and SMO engagement rates improving by 41.3%. The framework implements adaptive algorithms that automatically optimize keyword selection, content distribution strategies and social media engagement patterns through continuous learning mechanisms. Results indicate that the integrated approach significantly outperforms individual optimization strategies with the meta-level programming component reducing manual intervention by 67% while maintaining superior performance metrics. The research addresses critical gaps in current literature by providing a unified approach to digital marketing optimization that adapts to changing consumer behaviors and search engine algorithms. The framework's ability to process multi-dimensional marketing data and generate actionable insights in real-time represents a significant advancement in marketing automation technology. This study contributes to the field by demonstrating how meta-programming principles can be effectively applied to marketing optimization, providing both theoretical foundations and practical implementation guidelines for industry adoption.

Keywords: Meta-level programming, SEO optimization, Social media optimization, Digital marketing automation, Machine learning, Marketing analytics, Consumer behavior analysis, Performance optimization

Abstract: In the era of digital education, monitoring student engagement has become a critical factor in ensuring effective learning outcomes. This project presents a facial expression-based system designed to analyze and evaluate student engagement in online learning environments. By leveraging computer vision and deep learning techniques, specifically convolutional neural networks (CNNs), the system detects facial expressions and emotional cues through real-time video input from webcams. These expressions are then classified to determine levels of attentiveness, interest, and emotional state, which are key indicators of engagement.

Abstract: The rapid transformation of education into digital platforms has emphasized the need to improve virtual learning experiences by understanding students' emotions during lectures. Emotional states directly impact students’ focus, engagement, and learning outcomes, making real-time emotion analysis a valuable tool for enhancing teaching methodologies. This research presents an advanced emotion-based interactive dashboard designed to analyse students' facial expressions during online lectures, offering actionable insights to educators for improving teaching strategies and engagement. A key challenge in emotion recognition is dealing with occluded facial data caused by factors such as poor lighting, low resolution video, or face coverings. To address this, we employ a regenerative Generative Adversarial Network (GAN) capable of reconstructing occluded regions of the face while preserving critical emotional cues. The reconstructed data is processed using a deep learning model that predicts and classifies emotions into categories such as happiness, sadness, anger, surprise, fear, and neutrality. These emotional insights are then integrated into an intuitive dashboard that combines contextual data, such as the subject being taught, teaching faculty, and session-specific parameters. The dashboard offers dynamic visualization of emotion distribution, engagement trends, and real-time analytics, enabling educators to identify patterns in student behaviour. The system was validated using the CK+ dataset, achieving notable accuracy in classifying various emotions, even under conditions of partial facial occlusion. The integration of emotion-based analytics provides a unique approach to monitoring class engagement, identifying struggling students, and fostering personalized learning experiences. By combining advanced deep learning techniques with real-time analytics, the proposed system has the potential to redefine the future of online education, making it more responsive, adaptive, and student centered.

Keywords: Analytical Dashboard, Regenerative Generative Adversarial Network (GAN), Occluded facial data, Real time emotion.

Abstract: As the demand for energy-efficient electronic devices continues to grow, the focus on low-power memory designs becomes paramount. Our Project presents a comprehensive study on the design and verification of low-power SRAM memory cells, addressing the critical need for energy-efficient memory solutions in contemporary electronic systems. The primary objective is to optimize SRAM cells for reduced power consumption while maintaining satisfactory performance and reliability. The basic 6T Static Random Access memory (SRAM) cell experience from relatively high static and total power loss problem, to solve this 4T SRAM cell is designed. As the technology is shrinking, a significant amount of attention is being paid on the design of high stability Static Random Access (SRAM) cells in terms of static Noise Margin (SNM) for different levels of cache memories. This project presents a qualitative design of 4T Static Random Memory Access cell in terms of Read cell current, Write time, Static Noise Margin (Read and Hold), Write Noise Margin in 45nm and 90nm CMOS technology.

Keywords: SRAM, CMOS, EEPROM, PMO, NMO, PMI.

Abstract: Birds play a dual role in agriculture: while they contribute to the ecosystem by controlling pests, they can also cause significant crop damage. Traditional deterrents like scarecrows and sound devices often prove ineffective. This project presents a smart bird repellent system using image processing (MATLAB), deep learning (CNN), and IoT (Arduino, GSM). It detects harmful bird activity in real time and triggers a buzzer to scare them off. PIR sensors identify intruders, and a GSM module alerts the farmer. The system is low-cost, eco-friendly, and enhances agricultural productivity by reducing crop loss.

Keywords: Bird Detection, MATLAB, IoT, CNN, Arduino, PIR Sensor, GSM Module

Abstract: Liver diseases present significant diagnostic and management challenges due to their asymptomatic progression and the limitations of traditional diagnostic methods. Machine learning (ML) and deep learning (DL) techniques have emerged as transformative tools in liver disease diagnostics, enabling improved accuracy, efficiency, and automation in tasks such as disease classification, liver segmentation, and lesion detection. This review consolidates findings from recent studies, covering the use of logistic regression, support vector machines (SVMs), and convolutional neural networks (CNNs) in analysing clinical and imaging data. Advanced models such as DenseNet, YOLOv8, and DBN-DNN have demonstrated state-of-the-art performance in lesion detection, real-time diagnosis, and segmentation, achieving accuracy rates exceeding 95% in most cases. Despite their promise, challenges such as dataset limitations, variability in imaging protocols, and model interpretability remain significant barriers to clinical adoption. Future research should focus on enhancing generalizability across imaging modalities, incorporating explainable AI (XAI), and optimizing real-time deployment. This review highlights the potential of ML to revolutionize liver disease diagnostics, bridging existing gaps and paving the way for scalable, accurate, and efficient clinical solutions.

Keywords: Machine Learning, Liver Disease, Deep Learning, Segmentation, Classification, Real-Time Diagnostics, Multi-Modal Integration.

Abstract: Voice-controlled robotic systems have gained significant attention in recent years due to their ability to enhance automation and remote operation. This project presents a voice-controlled robot integrated with a camera module and an advanced hold and release mechanism, designed for efficient object manipulation. The system utilizes speech recognition technology, allowing users to command the robot hands-free. The incorporation of a high-resolution camera module enables real-time monitoring, visual tracking, and object identification, enhancing situational awareness and operational precision. The hold and release technology are a key feature that enables the robot to grip objects securely and release them with controlled precision, minimizing the risk of damage. This mechanism ensures adapt ability across diverse applications. An embedded micro controller governs real-time processing, interpreting voice commands, and executing tasks with accuracy. Furthermore, wireless connectivity allows remote access, facilitating seamless control and data transmission. This robotic system finds practical applications in several fields. The integration of a camera module makes the robot highly suitable for security and surveillance, where it can monitor restricted areas and detect potential threats. By combining voice recognition, computer vision, and precision handling, this robotic system represents a significant advancement in automation. Future enhancements may incorporate machine learning-based grasping techniques, and multi-language support to improve accessibility. With continuous development, such systems can make human-robot interactive more effective.

Keywords: IOT-based robot, Natural language processing (NLP), Bluetooth control robot, Wi-Fi surveillance robot.

Abstract: In this study, we explore the application of DNN algorithm for detection of fake-news, focusing on the role of attention mechanisms in improving performance. Four Algorithms were developed and assessed CNN, Bi-LSTM, Attention Convolutional Neural Network (ACNN), & Attention Bidirectional LSTM (ABiLSTM) to investigate their ability to accurately identified faked news by appropriate capturing context and semantic information in text. The LIAR database was applied to comprehensively assess the performing of such algorithm across training, validation, and test sets. Our results show that deep learning technique could improved the ability of deep models to focus on key components of the text, thereby improving such abilities to distinguish fake from true news. Among the models, the two attention-based methods, ACNN and ABiLSTM, demonstrated higher test accuracy of 0.56, reflecting a slight improvement over their non-attention counterparts. Furthermore, these models maintained a desirable balance between precision and recall, which underscores their robustness and ability to perform well across different evaluation criteria. Additionally, the F1-scores of the attention models were notably higher. Specifically, the ACNN and ABi-LSTM technique achieved F1-scores of 0.748 and 0.77 on the test set, outperforming the non-attention variants (CNN and BiLSTM). On the validation set, the F1-scores were 0.66 and 0.67, further validating their improved ability to extract and leverage important context-dependent features in the text. Among the two, ABiLSTM performed slightly better, suggesting that combining bidirectional LSTM with an attention mechanism is particularly effective for detecting fakes-news. Overall, this studies highlights the potential of attention mechanisms to enhance deep neural network models by focusing on the most informative components of text. The results underscore the importance of integrating attention into deep architectures to achieve greater robustness, accuracy, and generalization in detecting fakes news tasks.

Keywords: Bi-LSTM, Hybrid Model, Fake News

Abstract: The major problem faced by visually impaired people is that they are unable to do text recognition on their own, which forces them to depend on others for their day to day activities such as reading newspapers, letters sent through post, referring books etc. The aim of this project is to develop an Optical Character Recognition (OCR) system integrated into a college management system is designed to improve accessibility for visually impaired students, teachers, and administrators. This system uses advanced OCR technology, combined with image processing, Text-to-Speech Conversion (TTS) and machine learning algorithms, to convert printed text into a digital format. The key objective is to ensure that visually impaired users can access educational, administrative, and communicative resources seamlessly and can be recognized text into spoken words. For students, the OCR system allows them to access a variety of academic resources, such as textbooks, class notes, assignments, schedules, and exam papers, in an accessible digital format. The system can instantly extract text from documents and convert it into a format that is compatible with Text-to-Speech software, enabling students to interact with materials independently. Furthermore, students can benefit from multi-language support, ensuring they can access information in their preferred language. For teachers, the OCR system provides the ability to convert printed or handwritten teaching materials and notices into accessible formats for their visually impaired students. Teachers can scan documents and instantly create digital copies that are compatible with assistive technologies. This facilitates smoother distribution of teaching resources and more inclusive classroom engagement. Teachers can also use the system to modify documents to meet specific needs of students, enhancing personalized learning.

Keywords: OCR, TTS, Visual Impairment, Accessibility, Assistive Technology

Abstract: The utilization of deep learning and transfer learning methodologies in the field of image classification has led to substantial advancements across a range of applications. This research paper presents an evaluation of the VGG16 convolutional neural network, enhanced through transfer learning and regularization techniques, for the specific task of sheep breed classification. The model was fine-tuned on a dataset of six sheep breeds, using preprocessing techniques like resizing, normalization, and augmentation. Architectural enhancements such as batch normalization, dropout, and dense layers helped reduce overfitting and improve generalization. A dropout rate of 0.5 with a batch size of 16 achieved the highest test accuracy of 80.00%. Higher dropout rates (e.g., 0.8) resulted in underfitting and lower performance. Overall, the use of transfer learning and dropout regularization significantly improved classification accuracy.

Keywords: VGG16, Transfer learning, Regularization Techniques, CNN, Sheep Classification.

Abstract: Fungal skin infections are an emerging public health issue in India, with millions of people being affected every year by diseases like Tinea capitis, vaginal candidiasis, and aspergillosis. This review delves into the two sides of this challenge—evaluating the projected disease burden and discussing the emergence of Machine Learning (ML) and Deep Learning (DL) tools in fungal skin disease research during the period from 2018 to 2023. Epidemiological findings demonstrate a widespread incidence of superficial and systemic fungal infections, emphasizing the need for increased awareness, early detection, and efficient treatment approaches. At the same time, the review points to a significant rise in ML/DL-based research, indicating an intensifying interest in using artificial intelligence for dermatologic diagnosis. The convergence of public health and technology implies potential prospects for enhancing outcomes through AI-assisted tools, as long as they are supplemented by strong clinical validation and health policy infrastructure. The research calls for a multi-disciplinary solution to address India's increasing burden of fungal disease.

Keywords: Fungal skin diseases, India, Tinea capitis, vaginal candidiasis, machine learning, deep learning, dermatology, artificial intelligence, epidemiology, disease burden.

Abstract: Numerous methods, including electroencephalography (EEG) and magnetic resonance imaging (MRI), have been proposed to diagnose epileptic seizures. Deep knowledge (DL) is one of the many subfields of artificial intelligence. Conventional machine learning algorithms involving point birth were used prior to the emergence of DL. As a result, their performance was restricted to what the people creating the features by hand could do. However, in DL, the creation of features and type is completely automated. Similar to how the theory of epileptic seizures has advanced significantly, these methods have appeared in numerous medical fields. This study presents a thorough overview of a factory focused on automated epileptic seizure discovery using neuroimaging modalities and DL methods. Different approaches have been suggested to diagnose epilepsy.

Keywords: LSTM, EEG modalities, MRI modalities.

Abstract: The ability to predict channel capacity in wireless communication systems is critical for optimizing network performance and ensuring efficient data transmission. This work utilizes machine learning techniques to predict channel capacity based on key environmental and network parameters, including Signal-to-Noise Ratio (SNR), bandwidth, fading coefficients, and interference. Simulated data is generated to model the relationship between these factors and channel capacity using Shannon’s theorem. A Random Forest Regressor is employed to develop a predictive model, with hyper parameter tuning carried out using Grid Search CV for optimal performance. The model's performance is evaluated using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R². Visualizations are provided to illustrate the relationship between actual and predicted values of channel capacity, as well as the feature importance ranking. The work concludes with the saving of the trained model and scaler for future use. This predictive model serves as a step toward more intelligent and adaptive wireless network management, providing insights into optimizing communication systems under varying conditions.

Keywords: ML, MAE, MSE, RMSE, EDA.

Abstract: The Fish health is a critical factor in the success of aquaculture. Timely detection of diseases is essential to prevent the rapid spread of infections, minimize fish mortality, and reduce reliance on antibiotics. Traditional methods of disease detection rely heavily on manual inspection, which is time-consuming and prone to human error. This project proposes an automated system for fish disease detection using image-based machine learning techniques. By leveraging computer vision and deep learning algorithms, the system aims to efficiently identify diseases in fish through image analysis, offering a faster, more reliable alternative to traditional diagnostic methods. The system uses convolutional neural networks (CNNs) for classifying fish images into categories of healthy and diseased states. The dataset consists of fish images exhibiting various symptoms of diseases like white spot disease, fungal infections, fin rot, and bacterial gill disease. By training the CNN on these labeled images, the model can accurately predict the health status of fish in real-time, offering significant improvements in aquaculture management. The rapid growth of aquaculture as a global food production sector has increased the need for efficient and effective fish health management. Diseases in fish can lead to significant economic losses due to mass mortality, reduced production efficiency, and the use of antibiotics and other chemicals. Early and accurate detection of fish diseases is crucial for minimizing such risks. Traditional methods of disease diagnosis in aquaculture, which often rely on manual inspections by experts, are labor-intensive, time-consuming, and prone to errors. This study proposes a solution to automate and enhance the disease detection process through the use of image-based machine learning techniques, specifically employing deep learning algorithms like Convolutional Neural Networks (CNNs) for classifying fish diseases.

Keywords: Fish Disease Detection, CNN, Deep Learning, Aquaculture, Health Management.

Abstract: This research proposes a novel hotel recommendation system that addresses the limitations of single-criterion ratings by utilizing a multi-criteria collaborative filtering approach. The system integrates matrix factorization with a deep neural network to predict individual criteria ratings and employs Dempster-Shafer Theory (DST) to handle uncertainty in those predictions. By aggregating multiple ratings using evidential reasoning, the system provides a robust overall hotel recommendation. Experiments conducted on a real-world TripAdvisor dataset show the proposed method achieves superior accuracy compared to traditional and state-of-the-art models in terms of MAE, RMSE, and Coefficient of Determination.

Keywords: Hotel recommendation system, multi-criteria collaborative filtering, deep learning, matrix factorization, Dempster- Shafer theory, evidential reasoning.

Keywords: Women’s Safety, Real-Time Alert System, IoT- based Emergency Response, GPS Tracking, GSM Communication, Wearable Technology, Panic Detection, Embedded Systems.

Abstract: Indoor localization remains a challenging problem, as conventional GPS technologies often fail in enclosed spaces due to signal attenuation, multipath effects, and signal blockage caused by walls and other obstacles. This limitation poses significant hurdles in locations such as factories, hospitals, and warehouses, where accurate indoor tracking is essential. Overcoming these barriers is crucial for the advancement of numerous applications that depend on precise positioning, such as autonomous robotics, asset tracking, and navigation assistance within large facilities. In this project, we implement our solution within a controlled indoor lab setup that replicates real-world use cases, ensuring practical validation of our methodology.This research addresses these challenges by introducing a novel, cost-effective indoor positioning system (IPS) designed specifically for environments where traditional GPS cannot deliver reliable or accurate localization. The proposed IPS utilizes an array of ultrasonic sensors (HC-SR04) in conjunction with Time-of-Flight (ToF) measurements to calculate short-range distances with centimeter-level precision. The core system is built around an Arduino Mega microcontroller, which orchestrates real-time data acquisition and processing. The system architecture includes multiple anchor nodes and a mobile tag, coordinated to triangulate the tag’s position accurately. The integration of these off-the-shelf components ensures that the system remains scalable and affordable, making it accessible for widespread deployment in settings such as warehouses, hospitals, and smart buildings. Synchronization protocols and data filtering techniques are implemented to minimize measurement noise and environmental interferences, thereby enhancing the robustness and accuracy of position estimates.A Python-based visualization interface complements the hardware setup by providing a user-friendly platform for monitoring and managing localization data in real-time. Central to the software is a Python-implemented can , which calculates the two-dimensional position of the mobile node based on the distance measurements from multiple fixed anchor nodes. This algorithm processes the ToF data to derive accurate spatial coordinates, enabling real-time tracking of mobile agents. The novelty of our work lies in combining low-cost ultrasonic modules with custom-built synchronization, filtering mechanisms, and algorithmic positioning, achieving high precision without expensive hardware or complex infrastructure. The seamless coupling between hardware sensing and software visualization establishes the system as a practical and efficient solution for indoor navigation, offering a flexible foundation for further research and application development in the burgeoning field of smart environments.

Keywords: Indoor Positioning System (IPS), Ultrasonic Sensors, Time-of-Flight (ToF), Arduino Mega, Real-time Localization, Robotics, Automation, Smart Buildings, Asset Tracking, Trilateration Algorithm, Python.

Abstract: Diabetic Retinopathy (DR) represents a progressive ocular pathology characterized by retinal deterioration resulting from sustained hyperglycemia in diabetic patients. This microvascular complication constitutes the predominant etiology of visual impairment among working-age populations in developing nations. Given the irreversible nature of vision loss associated with advanced DR, therapeutic interventions primarily focus on preserving residual visual function through early detection and timely management. The current diagnostic paradigm relies heavily on manual interpretation of retinal fundus photography by ophthalmological specialists, creating significant challenges in terms of time consumption, economic burden, and resource allocation. These limitations are particularly pronounced during initial disease stages when pathological manifestations may be subtle and difficult to identify through conventional screening methods. Contemporary artificial intelligence approaches, specifically deep learning algorithms, offer promising solutions for automated analysis of retinal imagery, facilitating earlier diagnosis and more efficient screening protocols. This comprehensive review examines various automated methodologies developed for detecting DR and classifying its severity, providing a detailed analysis of their performance characteristics, dataset utilization, and clinical applicability. The investigation encompasses multiple deep learning architectures, their comparative advantages, and the potential for integrating them into existing healthcare delivery systems.

Keywords: Artificial Intelligence, Convolutional Neural Networks, Medical Image Analysis, Retinal Pathology, Computer-Aided Diagnosis

Abstract: In response to escalating cybersecurity threats, this paper explores an innovative framework combining team optimization algorithms and Convolutional Recurrent Neural Networks (CRNNs) to enhance cybersecurity measures. Traditional cybersecurity frameworks often face challenges in effectively handling the dynamic and complex nature of modern threats. By integrating team optimization algorithms, such as genetic algorithms and ant colony optimization, with CRNNs capable of processing both sequential and spatial data, this study proposes a robust solution. The framework aims to improve detection accuracy, response time, and overall resilience against sophisticated cyber attacks. Through empirical evaluations and case studies, we demonstrate the effectiveness and versatility of the proposed approach in real-world cybersecurity applications.

Keywords: Cybersecurity, team optimization algorithms, genetic algorithms, ant colony optimization, Convolutional Recurrent Neural Networks (CRNNs).

Abstract: Grading of pulses is a critical step in maintaining quality control within the agricultural and food processing industries. Pulses, which include various legumes like lentils, chickpeas, and beans, are widely consumed for their high nutritional value. Ensuring the quality of these products before they reach the consumer is essential. Traditionally, the grading process has been carried out manually by experts who visually inspect the pulses for features such as size, color, shape, and presence of defects. However, manual grading is often laborintensive, time-consuming, inconsistent, and susceptible to human error and fatigue. To overcome these limitations, the use of automated grading systems based on image processing techniques has gained significant attention. Image processing offers a non-invasive, efficient, and repeatable method for analyzing the physical characteristics of pulses. High-resolution images of the pulses are captured using cameras, and advanced digital image processing algorithms are applied to extract features such as area, aspect ratio, perimeter, color histogram, and surface texture. These features are then analyzed using rule-based systems or machine learning models to classify the pulses into different quality grades, commonly labeled as Grade A, B, and C. Grade A pulses typically exhibit uniform size, regular shape, consistent color, and minimal surface defects. Grade B may show minor irregularities, while Grade C usually includes broken or discolored grains and visible defects. Automated systems can be trained to recognize these patterns with high precision, reducing variability and enhancing the objectivity of the grading process. In addition to improving accuracy and consistency, automated pulse grading significantly reduces the time and manpower required for large-scale inspections. This is particularly beneficial for industries handling vast quantities of pulses where rapid and reliable quality assessment is crucial for productivity and profitability. Furthermore, digital records of graded batches can be maintained for traceability and quality auditing. The integration of image processing in pulse grading not only boosts operational efficiency but also supports farmers and suppliers by ensuring fair pricing based on product quality. It aligns with modern agricultural practices that emphasize technology-driven solutions for quality assurance and sustainability. In conclusion, automated grading of pulses using image processing is a transformative innovation in agri-tech. It addresses the limitations of manual grading by offering a faster, more consistent, and accurate method for quality assessment. As technology advances, such systems are expected to become more accessible and widely adopted across the pulse processing industry.

Abstract: In multilingual scene text recognition, accurate identification of the script used in each text region is essential before applying language-specific OCR. This paper proposes a lightweight script classification module based on MobileNetV2 [1], integrated into a broader Telugu scene text recognition pipeline. The system first detects word-level text regions using an enhanced EAST detector and then classifies each region into one of three script classes Telugu, English, or Hindi. The proposed classifier leverages transfer learning, efficient preprocessing, and a balanced dataset augmented to address class imbalance. Experimental results show that the classifier achieves a high overall accuracy of 94.81%, with minimal inter-script confusion, even in visually cluttered scenes. Qualitative examples and a detailed confusion matrix validate the model’s robustness and generalizability. This approach demonstrates how lightweight deep learning models can be effectively used in real-world OCR systems, particularly for Indian languages. Future directions include expanding script coverage, enabling handwritten text recognition, and integrating the module into an end-to-end OCR pipeline.

Keywords: Script Classification, MobileNetV2, Multilingual Scene Text, Transfer Learning, OCR Pipeline.

Abstract: Pollution attacks, such as transaction spam or fake data injection, undermine blockchain efficiency and security. This paper proposes a smart contract-based framework to autonomously detect and mitigate such attacks in real time. By embedding heuristic rules and anomaly detection logic into smart contracts, our system identifies malicious activity (e.g., excessive invalid transactions) and enforces countermeasures, including stake slashing or transaction throttling. Implemented on Ethereum, the solution demonstrates improved network resilience with minimal overhead, offering a decentralized and scalable defense against pollution threats while preserving data integrity.

Keywords: Pollution attacks, Smart contracts, Anomaly detection, Block chain security, Decentralized mitigation, Ethereum

Abstract: Artificial intelligence (AI) can be easily implemented in e-commerce and brought positive changes to the way that users interact with online platforms as a basis for shopping. This project involves the implementation of an industry-ready, deep learning-based recommendation system developed for e-commerce grocery applications with both content-based and collaborative filtering incorporated in a web-app framework written in Flask. The system uses content-based filtering by taking into consideration the metadata of the product, which are the description, category, nutritional information, etc., and users’ buying history, to provide recommendations in harmony with the users’ preferences. At the same time, collaborative filtering identifies patterns for the entire user base and uses additional methods, including matrix factorization and k-nearest neighbors that find the similarities between users and items, increase the range of recommendations and their relevance. For optimal usability, the recommendation system is embedded in a Flask web application, which offers a practical and hierarchical interface where the user can navigate through the grocery products list, state the preferences, and get the recommendations. The features of implementation include technical support for practical database, real-time computation of recommendations and actual scalability for the large amount of data. This sub-discipline deals with performance using algorithms, with help of Python tools – Scikit-learn and Pandas —logarithm for analyzing the data and accuracy is measured with the help of such basic metrics as precision, recall and RMSE. This integration approach links the concepts of AI and web development to redefine the conversation of grocery shopping from LSTM ensuring the users get accurate, timely and attractive product suggestions as a means of boosting the convenience and satisfaction.

Keywords: e-commerce grocery, content-based filtering, collaborative filtering, Flask web application, personalized suggestions, machine learning, user behavior analysis.

Abstract: The development of Large Language Models (LLMs) has significantly transformed the field of artificial intelligence, enabling machines to understand and generate human-like language. While most LLMs are trained on dominant global languages like English, there is a growing need to include regional languages such as Kannada to ensure linguistic inclusivity and cultural representation. This research focuses on the transformation and application of LLMs for the Kannada language. It explores data collection, preprocessing, tokenization, and fine-tuning strategies to adapt LLMs effectively. The study also addresses challenges such as limited datasets, script complexity, and semantic nuances unique to Kannada. By building or adapting Kannada LLMs, this work aims to enhance natural language processing (NLP) capabilities for Kannada speakers, supporting applications like translation, chatbots, sentiment analysis, and digital education. This transformation is a step towards democratizing AI access across linguistic boundaries.

Abstract: Skin diseases affect millions globally, posing screening challenges due to complex lesion characteristics and limited access to medical expertise. Traditional screening methods are time consuming, often requiring extensive laboratory testing. Deep learning and machine learning techniques have gained significant traction in recent years, serving as powerful tools in tackling complex problems, particularly in areas requiring substantial prior knowledge, such as biomedicine. With the challenge of inadequate medical resources, these methods have found impactful applications in disease screening, emerging as a pivotal research focus on dermatology. This project aims to develop an automated skin disease screening system using machine learning and deep learning techniques. The system is designed to accurately identify skin diseases, enhance early detection, address existing challenges in screening and ensure accessibility and affordability for all. This provides a concise review of the classification of skin diseases, leveraging Convolutional Neural Networks (CNN) and K-Nearest Neighbors (KNN) to analyse skin lesion characteristics and evaluate imaging technologies. By exploring the strengths of CNNs due to its high performance in image classification and feature extraction. KNN providing evidence by identifying similar images, making it an explainable AI model. This study presents an Evidence based screening system a virtual dermatology platform leveraging cutting-edge artificial intelligence and deep learning techniques for efficient skin disease classification. Using pre-trained models like GoogleNet, EfficientNet, ResNet, DenseNet, MobileNet and achieving a classification accuracy of 97% through EfficientNet. significantly reducing screening time and cost. The proposed system optimizes preprocessing, transfer learning, model training and cross-validation, significantly improving accuracy. The results highlight AI's potential to revolutionize dermatological screening, reducing costs and improving early detection.

Keywords: Convolutional Neural Network; K-Nearest Neighbors; Evidence based screening; EfficientNet;