VOLUME 14, ISSUE 10, OCTOBER 2025

Abstract: The foundation of this project lies in building a complete image classification system powered by Convolutional Neural Networks (CNNs), developed efficiently using TensorFlow and Keras. This approach aims to automate one of the most vital tasks in computer vision — categorizing visual data — with strong accuracy and reliability. The workflow begins with dataset preprocessing, which prepares the input images by normalizing pixel values to a fixed scale and resizing them into a consistent tensor shape, ensuring the CNN receives standardized data. The next key phase is data augmentation, where techniques such as image rotation, flipping, and scaling are applied to artificially expand the dataset. These transformations enhance model generalization and help minimize overfitting. Once the model achieves the desired performance, it is deployed as an interactive web app using Streamlit. This deployment converts the complex deep learning model into a user-friendly interface, enabling real-time image predictions and showcasing the high accuracy and efficiency of the developed system.

Keywords: Image Classification, Convolutional Neural Network, CNN, TensorFlow, Keras, Data Augmentation, Deep Learning, Streamlit, Python

Abstract: Knowledge graphs (KGs), which represent entities and their relationships in structured semantic networks, have been increasingly applied across a range of diseases, including thyroid disorders, cardiovascular conditions, and neurological disorders. Despite these advancements, current diagnostic methods often face challenges such as incomplete data integration, limited scalability, and reduced diagnostic accuracy. These limitations highlight the need for innovative approaches to address the complex pathogenesis of COVID-19.This review explores the integration of knowledge graphs with deep learning techniques for advancing COVID-19 research and diagnostics. Relevant COVID-19 datasets spanning viral characteristics, transmission patterns, clinical manifestations, and public health outcomes can be transformed into domain-specific knowledge maps that capture essential biomedical entities and their interconnections. By embedding these graphs into low-dimensional continuous vectors, semantic representations can be effectively utilized in deep learning frameworks. Such hybrid models hold promise for improving case prediction, identifying key disease indicators, and enhancing diagnostic accuracy. The fusion of KGs and deep learning not only offers novel insights into the underlying mechanisms of SARS-CoV-2 infection but also provides scalable solutions for real-world applications such as early detection, prognosis, and therapeutic target identification. Ultimately, this approach has the potential to strengthen evidence-based decision-making in pandemic management and contribute to global efforts in mitigating the impact of COVID-19.

Keywords: Knowledge Graph, Disease Prediction, Electronic Medical Records, COVID-19, Deep Learning

Abstract: Stress is a universal concept that concerns people of all ages, from young to old, and even all living things. It is important to detect it to protect itself from the negative effects of stress, which is intertwined with human life, or to benefit from its existence. In this study, a model based on fuzzy logic techniques which are some of the sub-branches of artificial intelligence has been developed using photographs containing facial expressions to assess the stress levels of individuals. Basically, the difference between two photographs of the individual was used in the model. Various fuzzy logic techniques which are Fuzzy C-Means (FCM) clustering, Adaptive Neuro-Fuzzy Inference System (ANFIS) and Fuzzy Inference System (FIS) are used in proposed cascade model. In the result of the model, the stress level of the individuals was included in one of the “None”, “Low”, “Moderate” and “High” levels. The fuzzy model correctly assessed approximately 70% of the dataset used.

Keywords: Artificial intelligence, fuzzy logic techniques, fuzzy inference system, image processing, stress assessment

Abstract: The stock market is one of the most dynamic and unpredictable financial systems, influenced by a wide range of economic, political, and psychological factors. Stock prices fluctuate primarily due to supply and demand, making it challenging to accurately forecast market movements. Traditional methods of analysis, such as fundamental and technical analysis, have been used extensively, but the rise of machine learning and deep learning techniques offers new opportunities for more precise predictions. This research focuses on the application of machine learning models, particularly Long Short-Term Memory (LSTM) networks, for predicting stock market trends. The study leverages historical stock price data, technical indicators, and sentiment analysis from news and social media to train and evaluate the LSTM model. Experimental results demonstrate that LSTM networks can capture complex temporal dependencies in stock price movements, offering improved prediction accuracy compared to conventional methods. Furthermore, the research explores the integration of ensemble models and hybrid approaches combining LSTM with other machine learning algorithms to enhance prediction reliability. The study also discusses challenges such as overfitting, data preprocessing, and feature selection, providing insights into practical implementation for real-world stock market forecasting. This approach can assist investors, financial analysts, and traders in making informed decisions, optimizing investment strategies, and minimizing financial risks.

Keywords: Stock Market Prediction, Machine Learning, Deep Learning, LSTM Networks, Financial Forecasting, Technical Analysis, Sentiment Analysis, Time Series Prediction, Investment Strategies, Ensemble Models, Hybrid Models.

Abstract: Chronic Kidney Disease (CKD) is a progressive medical condition characterized by the gradual loss of kidney function over time.This paper presents a machine learning–based approach to predict CKD using clinical . The study focuses on data preprocessing techniques, including handling missing values, feature scaling, and encoding categorical variables, to enhance model accuracy and reliability. Several machine learning algorithms, such as Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine, are implemented and evaluated using performance metrics like accuracy, precision, recall, and F1-score. Among these models, the Random Forest classifier demonstrates superior predictive performance, achieving high accuracy and robust generalization across test data. The experimental results suggest that the integration of machine learning techniques in healthcare can significantly assist medical practitioners in early CKD detection, risk stratification, and informed clinical decision-making. Furthermore, this study highlights the potential of artificial intelligence to transform traditional diagnostic procedures into data-driven, automated systems for improved healthcare delivery. and speed of disease diagnosis. The outcomes of this study highlight the potential of artificial intelligence (AI) in supporting data-driven healthcare solutions and enabling early intervention strategies for patients at risk of CKD

Keywords: Chronic Kidney Disease (CKD) Kidney function prediction Machine Learning (ML) Medical data analysis Disease classification Health informatics.

Abstract: Society today faces profound transformations driven by artificial intelligence (AI), encompassing economic shifts, social dynamics, and ethical dilemmas. While traditional frameworks struggle with the scale and speed of AI-driven changes, this study explores how AI can amplify productivity, innovation, and equity, or exacerbate inequalities and risks. The objectives are to identify key AI impacts on society, assess their implications, and address governance and ethical challenges. Employing a mixed-methods approach, including literature review, case analyses, and policy examination, we investigate techniques such as machine learning for economic forecasting, natural language processing for social interactions, and algorithmic decision-making in public services. This synthesis highlights AI's potential for societal enhancement while underscoring the need for balanced implementation.

Abstract: The Fishermen Maritime Border Control System was created as a safety and security measure to improve situational awareness at sea and stop illegal border crossings. To track the position of the vessel and the weather in real time, the system combines environmental sensors, motor control, alarm modules, and radio frequency transmission. To make sure fishing vessels stay inside approved fishing zones, their location is monitored using an RF transmitter–receiver combination. The motor is automatically turned off when a vessel breaches the predetermined border, and an LCD warning and buzzer give prompt alerts so that remedial action can be taken. Simultaneously, the device measures variables like temperature and humidity to conduct ongoing weather monitoring. In order to warn fishermen of possible dangers, alerts are produced if crucial thresholds are surpassed. The suggested approach reduces the possibility of illegal marine entrance, increases adherence to fishing laws, and protects fishermen from unfavorable sea conditions by fusing automated border enforcement with environmental monitoring.

Keywords: Maritime Control. RF transmitter, RF Receiver, Environmental Monitoring, Border Control

Abstract: With rapid urbanization and the rising number of vehicles, air pollution has become a significant environmental concern. The proposed Vehicle Emission Monitoring System provides an efficient IoT-based solution for real-time detection and analysis of vehicular emissions. Sensor modules continuously monitor the concentration of harmful gases such as CO, CO₂, and NOx, while data is transmitted to a centralized server for evaluation. There is an increasing need for automatic pollution detection systems integrated with automatic number plate recognition (ANPR) to identify and alert vehicle owners, thereby assisting government agencies in enforcing environmental regulations. The developed system automatically triggers alerts when emission levels exceed permissible limits, displaying information on an LCD and generating notifications through connected networks. This intelligent framework contributes to sustainable urban management by enabling proactive pollution control, improving environmental awareness, and supporting regulatory compliance.

Keywords: Embedded System, Gas Sensor, Number Plate Recognition

Abstract: This paper proposes and experimentally confirms a ready-for-production hybrid quantum-classical protocol for detection of fraudulent transactions under credit card and UPI payment methods. We create an Integrated Dataset consisting of 3,044,322 anonymized transactions from different public and institutional sources and suggest a modular quantum-inspired feature-engineering flow that generalizes a 14-dimensional raw feature vector to 268 engineered descriptors employing amplitude/phase encodings, entanglement-motivated pairwise interactions, as well as measurement-pro Quantum-inspired features are integrated with efficient classical learning algorithms (LightGBM, XGBoost, CatBoost, RandomForest, ExtraTrees) in a learned stacking ensemble we call quantum weighted ensemble. It was trained with stratified 5-fold cross-validation with SMOTE-aware rebal applied solely on folds to reduce leakage. In a reserved temporal holdout test set, proposed system achieves AUC = 0.9269, Precision = 0.92, Recall = 0.96, F1 = 0.94, and latency 2.7 ms per transaction in simulated production mode. In comparison with a classically tuned RandomForest Baseline (AUC # 0.8851), The hybrid system reduces false positives to as low as ~71.2 false negatives by "~74.8%, attaining estimated yearly savings for a 100M transaction operator of the magnitude of US$1.36M due to lowered loss as well as research expenses. Ablation Scientific research reveals that largest marginal gains are entanglement-related pairwise encodings due to measurement probabilistic descriptions. All results are significant at a 0.001 level (.p < 0.001). required no special quantum hardware (quantum encodings are classically calculatedRoute) but is designed to be portable to NISQ devices for additional improvements. We elaborate on operational limitations, moral implications (confidentiality, transparency, rectification of false positives), and specific next: kernel porting on quantum processors, federated quantum learning for multi-bank cooperative learning and adversarial robustness assessment. This research shows a practicable method for almost- Quantum concepts to concretely enhance financial fraud defenses while remaining executable now.

Keywords: Quantum Machine Learning, Fraud Detection, Hybrid Computing, Financial Analytics.

Abstract: This research presents an artificial-intelligence-based framework developed to perform automatic summarization and visualization of PDF files The proposed system combines state-of-the-art NLP algorithms with an intelligent image-generation module to create informative visual forms of textual data. The framework follows a modular structure that includes text extraction, summarization, adaptive complexity assessment, and visual rendering units, helping users grasp long documents more effectively. The backend component is implemented in the Python Flask framework, and lightweight web technologies are utilized in the frontend to maintain a responsive user interface. Experimental findings indicate that the suggested method greatly enhances access to information and decreases the effort and time needed for reviewing lengthy materials

Keywords: PDF Summarization, Natural Language Processing, Artificial Intelligence, Image Rendering, Adaptive Complexity, Flask.

Abstract: Phishing, a deceptive cyber-attack technique aimed at extracting confidential information, has evolved into one of the most significant threats in modern cyberspace. Traditional detection systems relying on static rules, blacklists, or manual inspection fail to recognize rapidly evolving and AI-generated phishing attempts. This research presents an Artificial Intelligence (AI)-driven phishing attack detection and prevention framework that integrates Natural Language Processing (NLP) with Deep Learning (DL) to enhance real-time recognition accuracy. The proposed hybrid model combines Bidirectional Encoder Representations from Transformers (BERT) for contextual text understanding and a Convolutional Neural Network (CNN) for URL pattern analysis. Additionally, the integration of Explainable AI (XAI) techniques such as LIME and SHAP provides interpretability for each classification decision. Through experimentation on benchmark datasets like PhishTank and Kaggle, the system achieved an overall performance accuracy of 96.4%. The model exhibits strong adaptability, continuous learning capabilities, and superior resilience against zero-day phishing threats. This study contributes to a transparent, adaptive, and intelligent defense framework for the next generation of cybersecurity systems. To address these limitations, this research proposes an AI-Powered Phishing Attack Detection and Prevention System that integrates Natural Language Processing (NLP) and Deep Learning (DL) for intelligent, adaptive, and explainable threat detection. The system employs BERT (Bidirectional Encoder Representations from Transformers) to analyze the semantic and contextual meaning of email or web content, while a Convolutional Neural Network (CNN) model examines the structural and lexical characteristics of URLs. By combining these two analytical perspectives, the model forms a hybrid AI engine that significantly enhances accuracy and resilience against zero-day phishing attacks. A key innovation of this work lies in its Explainable AI (XAI) component, which utilizes tools such as LIME and SHAP to interpret the model’s decisions. This transparency allows users and cybersecurity analysts to understand the reasoning behind each detection result, thereby improving trust and system reliability. The system also integrates a real-time browser extension and interactive web dashboard that proactively prevents users from accessing malicious domains and provides analytical visualizations of phishing trends.

Keywords: Artificial Intelligence, Cybersecurity, Phishing Detection, Deep Learning, NLP, Explainable AI, BERT, CNN

Abstract: The agricultural sector in small towns and rural areas often faces challenges such as limited market access, reliance on intermediaries, and inadequate digital infrastructure. Falaahar.com is an innovative Agri e-commerce platform designed to bridge the gap between small-scale farmers and consumers by providing a localized, user-friendly digital marketplace. The platform focuses on direct farm-to-consumer transactions, eliminating middlemen to ensure fair pricing and fresher produce. With features like mobile-first design, vernacular language support, secure payment gateways, and real-time order tracking, Falaahar.com aims to enhance inclusivity, transparency, and efficiency in agricultural supply chains. This platform not only empowers farmers with better market reach and control over their sales, but also promotes sustainability by encouraging the consumption of locally sourced produce. Future enhancements include mobile application development, AI-driven personalized recommendations, and integration with government schemes to further support rural entrepreneurship and digital literacy. Falaahar.com represents a significant step toward transforming agricultural commerce for small-town vendors and consumers in India.

Keywords: Agri E-commerce, Small-town Farmers, Digital Marketplace, Fresh Produce, Supply Chain, Mob ile-first Interface, Vernacular Language Support, Logistics Optimization, Rural Accessibility, AI/ML in Agriculture.

Abstract: The integration of advanced Artificial Intelligence (AI) and Machine Learning (ML) technologies is fundamentally transforming interactive entertainment by enabling the development of non-Player Characters (NPCs) that are dynamic and highly responsive to player actions and evolving game environments. This study explores various AI techniques used in NPC development, including Deep Learning, Reinforcement Learning, Natural Language Processing (NLP), and Behavioral Cloning. It examines how these technologies enable NPCs to exhibit sophisticated behaviors, learn from player interactions, and adapt to changing game conditions. Key findings are derived from a comparative case study of Grand Theft Auto V (GTA V) and Red Dead Redemption 2 (RDR2), which reveals a critical trade-off between scale (GTA V prioritizes scale) and depth (RDR2 emphasizes authenticity). The research concludes that hybrid AI architectures combining multiple machine learning approaches yield superior NPC performance, establishing a paradigm shift toward creating authentic virtual worlds that respond intelligently, remember meaningfully, and evolve naturally over time, thereby setting new standards for player immersion and engagement. The document also discusses the technical challenges involved in implementing AI-driven NPCs, such as computational complexity, real-time processing constraints, and maintaining behavioral consistency.

Abstract: The rapid expansion of online travel platforms has created a demand for intelligent recommendation systems capable of assisting travelers in selecting destinations, attractions, and activities that match their preferences. Traditional approaches, such as collaborative filtering and content-based filtering, each have significant limitations when applied in the tourism domain. Collaborative filtering often struggles with data sparsity and cold-start scenarios, while content-based filtering can result in overspecialization and reduced diversity of recommendations. These challenges highlight the need for a more robust solution that leverages the strengths of both approaches. To address this gap, we propose an adaptive hybrid travel and tourism recommendation system that integrates collaborative filtering using Singular Value Decomposition (SVD) with content-based methods based on textual and categorical attributes of destinations. A weighted fusion strategy is introduced to balance personalization with contextual relevance, thereby improving both recommendation accuracy and diversity. The system is evaluated using a synthesized dataset of tourist attractions and user ratings, with results showing that the hybrid approach significantly outperforms standalone models in terms of Precision@K, Recall@K, and NDCG@K. This research demonstrates the potential of hybridization for developing scalable, context-aware tourism recommender systems and offers a practical framework for deployment in real-world travel platforms.

Abstract: The modern healthcare industry faces a "data deluge," characterized by massive, heterogeneous information streams from Electronic Health Records (EHRs), high-throughput multi-omics experiments (genomics, proteomics), and real-time Internet of Things (IoT) monitoring devices. Traditional computational methods are inadequate to manage the scale defined by the Four V's: Volume, Velocity, Variety, and Veracity. This paper proposes the design of a Precision Healthcare Analytics Platform, a scalable Big Data architecture intended to systematically ingest, integrate, process, and analyze this complex data. The architecture leverages Hadoop Distributed File System (HDFS) for massive, fault-tolerant storage and Apache Spark for high-speed, distributed processing and Machine Learning (ML) capabilities. The core objective is to integrate siloed clinical data with biomolecular profiles, facilitating a critical paradigm shift from population-based generalized care to patient-specific personalized medicine. By employing advanced analytics, including Natural Language Processing (NLP) and predictive modeling, the platform aims to enhance clinical decision-making, improve public health surveillance, and substantially reduce operational costs.

Keywords: Big Data, Personalized Medicine, Precision Healthcare, Hadoop, Apache Spark, Multi-Omics, Predictive Analytics, FHIR, Clinical Decision Support (CDS).

Abstract: The evolution of Human-Computer Interaction (HCI) is increasingly focused on developing interfaces that offer more organic and intuitive methods of system control, moving beyond the limitations of conventional hardware. This paper presents a novel, dual-modality framework that functions as a virtual mouse and system controller by integrating real-time hand gesture analysis with voice command interpretation. The primary goal of this solution is to provide a fully contactless interaction method, thereby improving accessibility for users with motor impairments and offering enhanced convenience in hands-free operational scenarios, such as academic presentations or sterile work environments. Our system is developed in Python, employing OpenCV for video stream capture and the MediaPipe framework for high-fidelity hand and finger landmark detection, enabling precise cursor manipulation and control over system parameters like audio volume and screen brightness. Complementing this, a voice interface, created using the Eel library and Google Text-to-Speech (gTTS), processes verbal instructions to perform tasks such as launching software, initiating web searches, and querying system status. The fusion of these two modalities results in a highly responsive and user-centric interface that significantly advances the state of touch-free computing.

Keywords: Human-Computer Interaction (HCI), Gesture Recognition, Voice Command, Computer Vision, MediaPipe, Accessibility.

Abstract: The world's future relies on ensuring children are born without complications, yet many suffer from disorders like brain injury or stillbirth caused by fetal distress due to insufficient oxygen supply during delivery. Traditional Cardiotocography (CTG) is a widely used method to analyze the fetal heart rate (FHR) and the mother's uterine contractions. However, its interpretation is often subjective and dependent on clinician expertise, leading to inter- and intra-observer variation and high false-positive rates. This project addresses this critical gap by developing a more reliable, accurate, and automated system for monitoring fetal well-being using CTG data. The solution leverages machine learning (specifically, the Random Forest Classifier) and signal processing techniques to analyze key physiological parameters, aiming to reduce human error, standardize diagnosis, and provide real-time alerts for timely clinical interventions. Ultimately, this initiative seeks to enhance the quality and accessibility of prenatal care, contributing to better maternal and neonatal outcomes.

Keywords: Fetal Distress, Cardiotocography (CTG), Fetal Heart Rate (FHR), Machine Learning, Random Forest, Automation

Abstract: Traditional Intelligent Tutoring Systems (ITS) fail to scale due to static, pre-programmed pedagogy. This paper introduces the Reinforced-LLM Tutor (RLT), a novel architecture overcoming these limits. RLT synergistically integrates Large Language Models (LLMs), Reinforcement Learning (RL), and multi-agent systems. The framework features four modules: a Retrieval-Augmented Generation (RAG) Domain Knowledge Module to ensure factual accuracy, a Dynamic Student Model tracking cognitive/affective states, a Multi-Agent Pedagogical Core (Expert, Socratic, Motivational agents), and an Adaptive Policy Engine. This engine uses RL, modeled as an MDP, to learn an optimal teaching policy, creating a truly adaptive, self-improving tutor.

Keywords: Intelligent Tutoring Systems, Large Language Models, Reinforcement Learning, Personalized Learning, Multi-Agent Systems, Educational Technology, Gamification.

Abstract: Phishing attacks continue to pose a significant cybersecurity challenge worldwide. This study presents a robust, adaptive URL-based phishing detection framework that integrates Balanced Random Forest, and XGBoost within a soft-voting ensemble architecture. The model utilizes lexical, structural, and domain-level features extracted directly from URLs, allowing real-time prediction without relying on blacklists. Experimental evaluation achieved an accuracy of 93.29%, precision of 92.30%, recall of 95.26%, F1-score of 93.76%, and ROC-AUC of 0.982. The ensemble demonstrates strong adaptability in identifying zero-day phishing URLs and can be seamlessly deployed via Flask-based APIs and browser extensions.

Keywords: Phishing Detection, Machine Learning, Ensemble Learning, Cybersecurity, URL Features.

Abstract: As devices, cloud services, and critical systems connect, cyber threats are increasing. Cyber I am an intelligent security agent that learns, adapts, and defends itself in real time. Using machine learning and reinforcement learning, it detects suspicious activity and reacts instantly to limit damage. Unlike static systems, Cyber I continuously update from new attack patterns and real-time threat data, providing organisations with a dynamic and reliable defence against both known and emerging cyber risks.

Keywords: Artificial Intelligence (AI), Cybersecurity, Autonomous Cyber Defence, Real-Time Threat Detection, Self-Learning Systems, Intrusion Prevention.

Abstract: The Unified Payments Interface (UPI) is a revolutionary real-time payment system developed by the National Payments Corporation of India (NPCI) that has fundamentally transformed India's digital economy. This project report presents a comprehensive analysis of the UPI system, framed as a large-scale software engineering project, from its conceptualization and planning to its design, implementation, and real-world impact. The system was conceived to address the critical challenges of a cash-dominant economy, aiming to provide a secure, interoperable, and highly convenient mobile-first payment solution. The design is based on a sophisticated four-pillar, API-driven architecture that decouples user-facing applications from the core banking infrastructure, fostering a competitive and innovative ecosystem of Payment Service Providers (PSPs). This model facilitates seamless peer-to-peer (P2P) and person-to-merchant (P2M) transactions using simple identifiers like a Virtual Payment Address (VPA) or QR code, eliminating the need to share sensitive bank account details. Key features include the integration of multiple bank accounts into a single application and mandatory two-factor authentication via a UPI PIN for all transactions. The development and rollout of the system were guided by a hybrid methodology, combining a Waterfall approach for the stable core infrastructure with Agile principles for the rapidly evolving ecosystem of third-party applications. Rigorous integration, performance, and security testing were paramount to ensure the system's reliability and ability to scale to billions of monthly transactions. The results since its 2016 launch have been phenomenal, with UPI now accounting for the vast majority of digital transaction volumes in India.

Abstract: The abstract begins by highlighting how machine learning (ML) and artificial intelligence (AI) are transforming healthcare by creating intelligent applications that assist doctors in making better decisions. Among these applications is the Medicine Recommendation System (MRS), which is designed to suggest the most suitable medicine for a patient. Unlike traditional prescribing methods that rely mainly on a doctor’s experience, the MRS uses patient-specific information such as symptoms, medical history, and demographic details to make recommendations. The research focuses on designing, implementing, and evaluating this system. To do so, it collects and processes different types of data, including electronic health records (EHRs), laboratory test results, and drug interaction databases. These inputs ensure that the system not only considers the patient’s current condition but also past illnesses and possible side effects that may arise from combining different drugs. To analyze this information, several machine learning algorithms are applied. Decision trees and random forests are chosen for their interpretability and robustness; logistic regression is used as a simple baseline model; support vector machines (SVM) are tested for their strength in classification problems; and deep neural networks are explored for their ability to recognize complex patterns in medical data. Once trained, the system is able to generate personalized medicine recommendations for new patients. This has significant benefits: it can improve prescription accuracy, minimize adverse drug reactions, and provide valuable support to healthcare professionals. Instead of replacing doctors, the system works as a decision-support tool, offering evidence-based suggestions that doctors can review and approve. In conclusion, the abstract conveys that an ML-driven Medicine Recommendation System has the potential to make healthcare more reliable, safe, and tailored to individual patients, ultimately improving the overall quality of medical treatment.

Abstract: Self-driving cars, also referred to as autonomous vehicles (AVs), represent one of the most promising technological advancements of the 21st century. They combine artificial intelligence, computer vision, machine learning, and sensor fusion to navigate and operate without direct human intervention. The purpose of this research is to design and implement a cost-effective self-driving car prototype that can perform lane detection, obstacle avoidance, and path navigation using open-source technologies. This paper discusses the theoretical background, development methodology, experimental evaluation, results, and implications for future mobility systems. The proposed system utilizes a Raspberry Pi, camera module, and ultrasonic sensors for perception and control. Experimental results indicate lane detection accuracy of 95% and obstacle avoidance success of 90% in controlled environments. The research concludes that while current limitations prevent full autonomy, low-cost self-driving prototypes play an essential role in advancing autonomous vehicle education and research.

Abstract: Retail sales prediction plays a crucial role in effective inventory management, marketing strategy, and business profitability. This research focuses on predicting the sales of Big Mart outlets using various machine learning techniques. The dataset contains information on different products and store attributes. We compare algorithms such as Linear Regression, Random Forest, and XGBoost to determine the best-performing model for accurate sales forecasting. The results show that ensemble-based methods outperform traditional regression models in prediction accuracy.

Keywords: Sales Prediction, Big Mart, Machine Learning, Random Forest, Regression, Retail Analytics

Abstract: Cloud storage has become ubiquitous, yet users remain surprisingly vulnerable despite the sophisticated security measures that major providers have put in place. Most security breaches do not occur because the technology fails; rather, they result from human error, poor choices, incorrect system configurations, or a lack of understanding of legal requirements. This study examines the persistence of this gap and its implications for privacy and regulatory compliance. This paper analyzes three cloud storage types that are supported by all cloud providers: block, file, and object storage, which affect security outcomes differently. The analysis draws on real-world incidents rather than hypothetical scenarios. The Capital One breach, for example, illustrates how theoretical weaknesses can quickly become major disasters. By analyzing such cases alongside the technical distinctions between storage models, this study identifies where and why security systems most frequently fail. The findings reveal that while cloud providers have largely addressed the technical aspects of security, human and organizational factors remain problematic. This has important consequences for privacy protection and regulatory oversight in cloud environments. The research also evaluates emerging security approaches, such as Zero Trust Architecture and confidential computing, and emphasizes practical protective measures including client-side encryption, tokenization, and multi-factor authentication. The study provides detailed coverage of major compliance frameworks, including GDPR, HIPAA, and ISO/IEC 27018, offering implementable strategies for technical controls and regulatory adherence. This work aims to strengthen cloud storage security by focusing on actionable privacy safeguards, deployable technical solutions, and compliance strategies that can be realistically adopted by users. The results should prove valuable for researchers studying cloud security, IT professionals designing storage systems, and policymakers developing data protection regulations in an increasingly digital world.

Keywords: Cloud computing, data privacy, data security, cloud storage services.

Abstract: Small and medium-sized enterprises (SMEs) often face challenges in managing inventory due to fluctuating demand, limited analytical resources, and manual tracking systems. These inefficiencies lead to frequent stockouts, overstocking, and financial losses. This paper presents an AI-driven inventory management system designed to leverage machine learning for demand forecasting, automate replenishment, and optimize stock levels. The proposed system integrates predictive analytics with real-time inventory tracking to support data-driven decision-making. Preliminary results demonstrate improved forecasting accuracy and enhanced operational efficiency, contributing to sustainable and intelligent business management.

Keywords: AI-driven inventory management, demand forecasting, machine learning, real-time tracking

Abstract: Efficient crop selection plays a crucial role in enhancing agricultural productivity and sustainability. Traditional farming practices often rely on farmers’ experience and general guidelines, which may not consider local soil characteristics and environmental variations. This study proposes a soil-based crop recommendation system using machine learning techniques to support data-driven agricultural decisions. The system utilizes soil parameters such as pH, nitrogen (N), phosphorus (P), potassium (K), temperature, humidity, and rainfall to predict the most suitable crop for a given region. A dataset comprising soil and environmental attributes was preprocessed and analyzed to train various classification models, including Decision Tree, Random Forest, Support Vector Machine (SVM), and Gradient Boosting algorithms. Performance evaluation based on accuracy, precision, recall, and F1-score demonstrates that ensemble learning methods outperform traditional classifiers. The proposed model provides a reliable, scalable, and user-friendly solution for optimizing crop selection, improving yield, and promoting sustainable agricultural practices. Future work includes integrating real-time IoT sensor data and satellite imagery for dynamic recommendations.

Keywords: Crop recommendation, machine learning, soil analysis, precision agriculture, Random Forest, data-driven farming, sustainable agriculture, decision support system. .

Abstract: Due to the significant advantages, it offers as a pure, renewable energy that is free of issues related to waste and fuel, there has been a growing interest in using solar energy to generate electric power using photovoltaic panels over the course of the past few decades. There is a need for smart controllers to convert the direct current generated from photovoltaic cells towards alternating current through inverters, the most prominent of which are the Model predictive control (MPC). MPCs are adaptable devices with multivariable control methodology provide best powerful execution than direct control,In this paper we propose an efficient controller for adjusting Photovoltaic micro-grid power system using ANN technique to overcome problems of floating energy and enhancing the overall all system performance efficiency as a proposed technique. The suggested model should have the abilities to cancelling the losses and disadvantages effect provided by LC filters and other limitations in controlling techniques. The required software which have been chosen to employ the proposed technique relays upon utilizing MatLab2020 simulation program.

Abstract: The AI Code Analyzer Agent is a full-stack web application developed using the MERN stack (MongoDB, Express.js, React.js, Node.js) integrated with Google Gemini, an advanced generative AI model. This system acts as an intelligent assistant that analyzes user-submitted code, identifies potential syntax and logical errors, and provides improvement suggestions. The AI feedback is formatted and rendered using Markdown for readability, while syntax highlighting is achieved using PrismJS for enhanced visual clarity. The system architecture includes a responsive frontend, secure backend communication, and a scalable database design. The application allows users to interact with AI in real time, receiving educational and practical insights to improve their coding skills. The project showcases how artificial intelligence can be integrated into traditional web-based development workflows to create intelligent, user-friendly, and scalable solutions.

Keywords: AI Code Review, MERN Stack, Google Gemini, Syntax Highlighting, Code Optimization.

Abstract: Cyberbullying detection using Natural Language Processing (NLP) aims to identify harmful or abusive content on online platforms. This research focuses on classifying text data into cyberbullying and non-cyberbullying categories using advanced NLP and machine learning models. The dataset includes a variety of online comments, which are cleaned, tokenized, and vectorized using TF-IDF techniques. Machine learning algorithms such as Logistic Regression, Random Forest, and Support Vector Machine are evaluated for performance. Results show that ensemble-based methods outperform simple classifiers, achieving high accuracy and precision in detecting cyberbullying content.

Keywords: NLP, Cyberbullying Detection, Text Classification, Machine Learning, Sentiment Analysis

Abstract: The integration of Internet of Things (IoT) technologies has transformed modern agriculture by offering automation, accurate control, and remote monitoring. This work introduces an IoT-enabled hydroponic monitoring and management system that enhances the efficiency and sustainability of soilless cultivation. The system utilizes an ESP32 microcontroller with built-in Wi-Fi to link to the Blynk IoT platform, facilitating live tracking and control through a mobile interface. Core environmental parameters such as moisture and pH are continuously measured using appropriate sensors. The collected data are processed by the ESP32 and transmitted to the Blynk application, where users can view real-time readings and perform necessary adjustments. A relay-based circuit automatically manages water and nutrient pumps to ensure timely and precise delivery to plants. This reduces manual intervention, conserves resources, and maintains stable growth conditions. The proposed system provides a low-cost, scalable, and user-friendly solution suitable for home gardening, academic experiments, and commercial hydroponic farms, contributing to sustainable and productive agriculture.

Abstract: The exponential growth in SIM card fraud incidents poses significant challenges to telecommunications security, resulting in substantial financial losses and identity theft cases worldwide. This paper proposes a novel hybrid machine learning framework that integrates rule- based filtering with Random Forest classification for effective SIM card fraud detection. The system analyzes four key behavioral parameters: IMEI change frequency, geographical mobility patterns, call activities, and SMS usage behavior. Our approach implements a multi-layer detection architecture that combines the transparency of rule-based systems with the pattern recognition capabilities of machine learning. The framework features an interactive Streamlit- based dashboard providing real-time monitoring, explainable AI insights, and comprehensive analytics. Experimental results demonstrate 92.5% detection accuracy with 85.7% recall rate and processing times under 5 seconds. The proposed solution addresses critical limitations of existing systems and offers a practical, scalable approach for telecom security applications, particularly in the Indian telecommunications context. The system combines rule-based filtering with Random Forest classification to analyze SIM usage patterns including IMEI changes, location behavior, call frequency, and SMS activity. It detects various fraud types such as SIM swapping, cloning, multiple activations, and abnormal usage patterns. Implemented with Python and Streamlit, the solution provides an interactive dashboard for fraud analysis, feature importance visualization, and risk scoring. The model achieves high accuracy in classifying fraudulent SIM cards while maintaining explainability through transparent decision-making processes. This project offers a practical, scalable solution for telecom companies and financial institutions to combat SIM-based fraud, enhancing security in the rapidly evolving digital landscape.

Keywords: SIM Card Fraud, Machine Learning, Hybrid Detection, Random Forest, Explainable AI, Telecommunications Security

Abstract: Air pollution is a major global issue affecting the environment and human health. With the increasing concentration of harmful gases and particulate matter, monitoring and predicting air quality has become essential. This research aims to design and implement an Air Quality Prediction System using Python and Machine Learning (ML) techniques. The system predicts the Air Quality Index (AQI) based on environmental features such as PM2.5, PM10, CO, NO₂, SO₂, O₃, temperature, and humidity. Machine learning algorithms such as Linear Regression, Random Forest Regressor, and Support Vector Regressor (SVR) are applied to build predictive models. The model is trained and evaluated using Python libraries like pandas, numpy, scikit-learn, and matplotlib. Experimental results show that the Random Forest algorithm gives the best performance with higher accuracy and minimal prediction error. The proposed system can assist government agencies, researchers, and the public in understanding pollution trends and taking preventive measures to improve air quality.

Abstract: Brain tumor detection is a critical task in medical imaging that requires timely and accurate diagnosis for effective treatment. Manual interpretation of MRI scans is time-consuming and prone to human error. Machine Learning (ML) and Deep Learning (DL) techniques have demonstrated re- markable performance in automating tumor detection, segmentation, and classification. This review paper provides a comprehensive overview of various ML methods applied to brain tumor detection, discusses datasets, algorithms, evaluation metrics, and highlights recent trends and future research directions. The paper aims to provide a clear understanding of the current state-of-the-art approaches and the challenges that remain in this domain.

Keywords: Brain Tumor, Machine Learning, Deep Learning, MRI, Image Processing

Abstract: Correct car price prediction is significant in automotive manufacturing as it offers important benefits to the producers, dealers, customer by supporting honest pricing, inventory management, and informed management. This research develops a machine learning system that predicts the prices of cars based on an alteration of attributes, like name, price, model, year, km driven, and type of fuel. The work used a large dataset; the dataset was filtered so that missing values, data inconsistencies, and outliers in the data were reduced. like Linear Regression, Decision Trees, and Random Forests are working to make predictive models. The presentation of these models is evaluated using metrics like R-squared R² for accuracy and reliability. The outputs display the ability of Machine Learning techniques to deliver more accuracy in car price predictions,s howing their practical stability in the automotive domain. By dealing with issues like data quality, feature selection, and model interpretability, this study offers a solid basis for developing similar predictive systems. Besides, the study suggests that improvement, including incorporating real-time market data, can be considered to increase the accuracy of the prediction. This study has made it clear that Machine Learning plays a very complex role in changing the pricing strategies and supporting the stakeholders in driving an active automotive market.

Keywords: Car price prediction, machine learning, Automotive Manufacturing, predictive modeling, Linear Regression, data preprocessing, feature selection, R-squared, pricing strategies, integration, model Real-time Market Data, inventory Management

Abstract: Since coronavirus has shown up, inaccessibility of legitimate clinical resources is at its peak, like the shortage of specialists and healthcare workers, lack of proper equipment and medicines etc. The entire medical fraternity is in distress, which results in numerous individual’s demise. Due to unavailability, individuals started taking medication independently without appropriate consultation, making the health condition worse than usual. As of late, machine learning has been valuable in numerous applications, and there is an increase in innovative work for automation. This paper intends to present a drug recommender system that can drastically reduce specialists heap. In this research, we build a medicine recommendation system that uses patient reviews to predict the sentiment using various vectorization processes like Bow, TF-IDF, Word2Vec, and Manual Feature Analysis, which can help recommend the top drug for a given disease by different classification algorithms. The predicted sentiments were evaluated by precision, recall, f1score, accuracy, and AUC score. The results show that classifier LinearSVC using TF-IDF vectorization outperforms all other models with 93% accuracy.

Keywords: Index Terms—Drug, Recommender System, Machine Learn- ing, NLP, Smote, Bow, TF-IDF, Word2Vec, Sentiment analysis

Abstract: Emotion detection plays a vital role in enhancing human–computer interaction by enabling systems to recognize and respond to human emotions. This project focuses on the development of an automated emotion detection system using Convolutional Neural Networks (CNNs), a powerful deep learning architecture for image-based pattern recognition. The system is trained on facial expression datasets such as FER-2013 or CK+ that contain labeled images representing emotions like happy, sad, angry, fear, surprise, disgust, and neutral. The images are preprocessed through grayscale conversion, normalization, and data augmentation to improve model performance and generalization. A CNN model is designed and trained to extract hierarchical features from facial images, followed by fully connected layers for emotion classification. The model’s performance is evaluated using metrics such as accuracy, precision, recall, and confusion matrix analysis. Once trained, the model is deployed using a web interface or real-time video stream to detect emotions from live webcam input. The results demonstrate that the CNN-based system achieves high accuracy in identifying human emotions and performs effectively in real-world scenarios. This project highlights the capability of deep learning techniques to build robust emotion recognition systems applicable in domains such as human–computer interaction, mental health monitoring, and smart surveillance.

Abstract: The AI-Driven Emergency Response System enhances emergency management by integrating artificial intelligence, real-time data analysis, and automation. The system uses machine learning, computer vision, and natural language processing to detect and classify incidents such as accidents, fires, or medical emergencies from various data sources including CCTV, IoT sensors, and public reports. It automatically alerts the nearest response units and optimizes routes using GPS to ensure rapid assistance. A centralized dashboard provides real-time monitoring and predictive insights for authorities. This AI-based framework reduces human error, shortens response time, and supports the development of safer and smarter cities.

Keywords: Artificial Intelligence, Emergency Response, Machine Learning, Smart City, Automation.

Abstract: NutriAI is a smart nutrition assistant that helps people get personalized food suggestions based on their health and dietary needs, especially focusing on Indian food. It uses artificial intelligence to understand individual preferences, health conditions, and food habits to recommend the best meal plans. This project aims to build an easy-to-use and affordable system that can suggest balanced meals, taking into account traditional Indian food choices. The paper explains the background, how the system is designed, tested, and what benefits it provides. NutriAI collects user data and uses a database of Indian foods to give accurate nutrition advice, helping people eat healthier. Although it is not a complete health solution, NutriAI is a useful tool for improving nutrition awareness and personalized diet planning.

Abstract: In modern software development, multiple developers often work simultaneously on the same project files. Managing changes, tracking modifications, and maintaining different versions of code becomes extremely challenging without an efficient control mechanism. Version Control Systems (VCS) have become an essential part of collaborative development, enabling teams to work together effectively and track the complete history of their codebase. The project “Building a Version Control System” aims to design and develop a simplified and educational version of a VCS that allows users to manage source code versions efficiently. It provides functionalities such as file tracking, version history, branching, merging, and rollback. The goal is to offer an intuitive and lightweight system suitable for individuals, small teams, and educational use.

Keywords: Version Control System, Distributed Systems, Source Code Management, Commit Tracking, Branching, Merging, Rollback

Abstract: The rapid evolution of Generative Artificial Intelligence (AI) has transformed the technological landscape by enabling automated creation across text, image, audio, video, and code domains. This research paper titled “Generative AI Tools and Platforms Landscape” presents a comprehensive analysis of current generative AI platforms, focusing on their technical capabilities, architecture, and application diversity. The study uses a data-centric and code-based approach, employing Python-based libraries such as Pandas, NumPy, Matplotlib, Seaborn, and Scikit-learn to preprocess, analyze, and visualize real-world data from generative AI tool repositories. The methodology involves systematic data cleaning, exploratory data analysis (EDA), and predictive modeling using Logistic Regression within a machine learning pipeline. Results indicate that multimodal platforms and open-source models exhibit stronger adaptability and innovation potential. Statistical visualizations, heatmaps, and correlation analyses reveal significant patterns among platform features, release trends, and modality diversity. This study contributes to understanding the evolving Generative AI ecosystem, offering insights into its current landscape and identifying potential research gaps for future development. The outcomes demonstrate the significance of open innovation, ethical governance, and model transparency in shaping next-generation AI platforms.

Keywords: Generative AI, Machine Learning, AI Platforms, Multimodal Models, Data Analysis, Logistic Regression, Open Source AI, Foundation Models, Artificial Intelligence Tools, Code-based Research.

Abstract: This study develops and compares machine-learning models to predict stress levels among MCA students under Pune University (SPPU) using a questionnaire-based dataset collected via Google Forms. The survey included 1000 responses covering demographics, academic, lifestyle, social, and personal factors. After preprocessing (cleaning, one-hot encoding, scaling), dimensionality reduction (PCA), and feature selection, four models were trained and evaluated: XG Boost, Random Forest, Principal Component Analysis + Support Vector Machine, and Logistic Regression. Models were assessed using accuracy, precision, recall, F1-score, confusion matrices, and ROC-AUC. Key predictors included sleep quality, family support, financial concerns, academic workload, and peer pressure. Among these XG Boost showed the best performance based on weighted F1-score and balanced accuracy. The findings provide insights for early stress interventions and student wellbeing programs.

Keywords: Student stress, mental health prediction, XG Boost, Random Forest, PCA, SVM, Logistic Regression, one-hot encoding, survey data, MCA students.

Abstract: This document presents the design and development of an AI-powered Customer Relationship Management (CRM) system aimed at automating B2B campaign analysis and lead optimization. The system integrates Artificial Intelligence (AI) and Machine Learning (ML) algorithms with traditional CRM functions to deliver automated campaign insights, lead scoring, and performance summaries. The platform enables marketing teams to create, monitor, and evaluate marketing campaigns, track engagement metrics such as email delivery and open rates, and manage leads from multiple data sources. AI algorithms process uploaded CSV data to produce automated insights, including lead quality, campaign reach, and engagement patterns. A real-time dashboard visualizes campaign progress and generates AI-based summaries to assist in strategic decision-making. Additionally, the system supports asynchronous task processing using Celery and Redis, enabling seamless report generation and data updates. This study demonstrates how AI-driven automation in CRM systems enhances data accuracy, reduces manual analysis efforts, and improves decision efficiency in B2B marketing.

Keywords: AI-Powered CRM, Machine Learning, Campaign Management, Lead Optimization, Automated Reporting, Django, Data Analytics

Abstract: Despite rapid progress in artificial intelligence, communication between hearing and non-hearing individuals still faces significant challenges. This research proposes a real- time sign language translation system that converts hand gestures into readable text using a hybrid YOLOv5-Mediapipe-PyTorch architecture. The framework leverages NVIDIA CUDA for accelerated inference and OpenCV for image preprocessing and display. The convolutional model is trained through transfer learning on a curated Indian Sign Language (ISL) dataset containing 26 alphabetic and multiple word-level gestures. The developed system achieves 96.2 % accuracy and functions in real time on standard GPU hardware. Translated text is rendered as live subtitles via OBS virtual camera, enabling accessibility on conferencing platforms such as Google Meet and Microsoft Teams. Experimental evaluation confirms that the YOLOv5–Mediapipe hybrid substantially reduces latency while maintaining high precision. This work demonstrates a scalable path toward inclusive communication technology bridging the gap between hearing-impaired and non- sign-language users.

Keywords: Sign Language Recognition, Deep Learning, YOLOv5, Mediapipe, CUDA, PyTorch, Real-Time Translation.

Keywords: PHP, MySQL, Web Application, Automation, Database Management, XAMPP Server.

Abstract: This review presents an overview of artificial intelligence–driven approaches for medicinal plant identification using leaf image analysis. It focuses on advanced techniques such as hybrid convolutional neural network (CNN) models, transfer learning, feature extraction, and image preprocessing. The paper also summarizes key datasets, performance metrics, and real-time frameworks including Flask and YOLO. Despite notable progress, challenges persist in dataset diversity, integration of phytochemical information, and practical deployment. Overall, AI-based models show strong potential to improve the accuracy, efficiency, and accessibility of medicinal plant recognition, contributing to sustainable healthcare and biodiversity preservation.

Keywords: Medicinal plant identification, Artificial Intelligence, Deep Learning, CNN, Image recognition, Transfer Learning, Feature Extraction, Flask, YOLO.

Abstract: Agriculture is the backbone of the global economy, providing food security and livelihoods to billions. However, traditional farming practices are facing severe challenges such as climate change, resource depletion, and labor shortages. Artificial Intelligence (AI) offers transformative potential to revolutionize agriculture through automation, data-driven decision-making, and precision management. This research explores the integration of AI technologies—such as machine learning, computer vision, Internet of Things (IoT), and robotics—into agricultural systems. The study emphasizes how AI can optimize crop yield prediction, pest detection, irrigation management, and soil monitoring. The findings suggest that AI-driven solutions can enhance productivity by 25–30%, reduce input costs, and promote sustainable farming. The paper concludes that the “AI-based farming revolution” is key to achieving smart, sustainable, and resilient agriculture for the future.

Keywords: Crop Recommendation: Suggests suitable crops based on soil and weather data.

Abstract: Breast cancer continues to be a major global health concern, with survival prediction being a key element in improving treatment outcomes and clinical decision-making. This study applies machine learning (ML) techniques to the METABRIC (Molecular Taxonomy of Breast Cancer International Consortium) dataset to classify patient Overall Survival Status as either Living or Deceased. Five ML algorithms—Logistic Regression, Naïve Bayes, Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbours (KNN)—are implemented after comprehensive preprocessing, including handling missing values, categorical encoding, and feature scaling. Model performance is evaluated using accuracy, precision, recall, F1-score, and confusion matrix. Results indicate that Logistic Regression achieved the highest accuracy (97.3%), closely followed by Random Forest and Naïve Bayes. The findings demonstrate the potential of ML techniques in assisting oncologists with survival prediction, offering a foundation for future integration into personalized medicine.

Keywords: Breast Cancer, Machine Learning, METABRIC, Survival Prediction, Logistic Regression, Random Forest.

Abstract: The AgreSense system is an intelligent agricultural solution that leverages a combination of the Internet of Things (IoT), artificial intelligence (AI), and advanced sensors for real-time monitoring of soil and crop health. It gathers critical data points, including soil moisture, temperature, pH, and electrical conductivity (EC). This information is then processed to deliver automated recommendations for optimizing irrigation schedules and nutrient management. Architecturally, the system is built using cost-effective sensors and ESP32 microcontrollers. It employs LoRa (Long Range) communication technology for efficient, long-distance data transmission. At its core, data analytics and machine learning models are used to process the incoming data, enabling predictive insights for irrigation needs and strategies to maximize crop yield. This paper outlines the design, practical implementation, and performance evaluation of AgreSense, focusing on its affordability, scalability, and overall contribution to the advancement of precision farming.

Keywords: - Smart Agriculture, IoT, LoRa, ESP32, AI, Machine Learning, Precision Farming, Soil Monitoring.

Abstract: Most farmers in india face the problem of not being able to reach industries directly and, therefore, are paid belatedly and inconsistently. The smart farmer-industry link system is a web-based system exclusively aimed at bridging this gap by allowing farmers to upload crop details and industries to directly request and negotiate the produce. It is integrated with real-time notifications, secure authentication, and an admin dashboard for activity monitoring. Further, it will be developed using html, css, and javascript; thus, it has a user-friendly interface that guarantees reliability in communication and further ensures transparency in transactions. This paper presents the system architecture, implementation details, results, and future enhancements toward promoting digital agriculture and efficient farmer-industry collaboration. Index terms: smart agriculture, farmer-industry linkage, web application, digital marketplace, crop management.

Abstract: This study presents the design and implementation of an intelligent chatbot for a luxury jewellery e-commerce platform. The system employs machine learning techniques to classify customer queries and generate contextually accurate responses. It applies text preprocessing methods such as normalization and feature extraction using vectorization techniques to convert text into numerical form. A deep learning model built with TensorFlow/Keras is used for intent classification, while a regression-based approach supports dynamic jewelry price prediction. The system also integrates fallback mechanisms and heuristic rules to ensure reliability, contextual consistency, and enhanced user interaction quality.

Keywords: Machine Learning, Deep Learning, Intent Classification, Natural Language Processing (NLP), TensorFlow/Keras, TF-IDF Vectorization, Text Preprocessing.

Abstract: The growing incidence of street crimes such as theft, robbery, assault, harassment, and illegal weapon possession has highlighted the urgent need for real-time surveillance and rapid response systems. This paper proposes a mobile-based crime detection framework that transforms everyday smartphones into intelligent CCTV devices capable of identifying suspicious activity. Leveraging deep learning techniques, the system detects violent behavior, identifies weapons (e.g., knives and firearms), and performs facial recognition to detect known suspects. Upon threat detection, alerts are sent instantly to authorities along with supporting evidence such as images, timestamps, and location. The proposed system offers a scalable, affordable, and effec- tive solution for urban surveillance—especially in areas lacking traditional CCTV infrastructure—thus enabling quicker law enforcement response and contributing to safer, smarter cities.

Keywords: Crime Detection, Mobile Surveillance, Weapon Detection, Violence Recognition, Real-Time Alert System.

Abstract: Database security is a critical aspect of information technology that ensures the confidentiality, integrity, and availability of data stored in databases. With the rapid growth of digital data, databases have become prime targets for malicious attacks, unauthorized access, and data breaches. This paper provides an overview of key database security concepts, common vulnerabilities, and modern solutions to safeguard data assets. Techniques such as encryption, authentication, and access control are discussed along siderecent advancements in anomaly detection and security auditing. The study emphasizes the importance of proactive security policies,continuous monitoring, and the implementation of database firewalls to prevent data leaks and misuse.

Keywords: Database Security, Data Integrity, Access Control, Encryption, Authentication, Cyber security

Abstract: Now a days Electronic Voting system is very popular system to collect the votes. E- Voting system is very popular but there is low security methods so to overcome this , We have proposed Advance Voting System where the high security verification applied , it verifies the user Phone Number, E-Mail Address and captures live image for verification purpose so that it can cancel out the proxy voting. This paper mainly focuses on Biometric security mechanism to overcome the drawbacks. Elections are a key part of democracy, but traditional voting systems often face problems like voter impersonation, multiple voting, ballot tampering, and slow counting. This study explores an Advanced Voting System using Biometric Verification to solve these issues. The system uses unique biometric traits such as fingerprints and facial recognition to verify voter identities, ensuring that each person can vote only once. Votes are securely stored in a digital database and counted automatically, reducing human errors and speeding up the election process. Testing of the system shows that it is accurate, efficient, and reliable. By combining technology with secure voting practices, this system can increase transparency, prevent fraud, and build greater public trust in election results. The study highlights the potential of biometric technology to make elections safer, faster, and more trustworthy, contributing to stronger democratic processes.

Keywords: Proxy Voting Prevention, Transparent Election Process, Biometric Verification, Election Security, Voter Authentication, Fraud Prevention.

Abstract: Quantum computing has emerged as one of the most revolutionary paradigms in computer science and physics. It leverages the laws of quantum mechanics—superposition, entanglement, and interference—to process information in ways that classical computers cannot. This article explores the fundamental principles, hardware architectures, quantum algorithms, and the challenges faced in realizing scalable quantum systems. Additionally, it discusses real-world applications in cryptography, machine learning, materials science, and communication networks. The paper concludes with an outlook on how quantum computing is shaping the next technological revolution.

Keywords: Quantum Computing, Qubit, Superposition, Entanglement, Quantum Algorithms, Cryptography, Quantum Internet, Quantum Supremacy, Quantum Error Correction.

Abstract: Agriculture remains the backbone of many economies, yet farmers continue to face major challenges such as unpredictable weather, pest infestations, crop diseases, and fluctuating market prices. This paper presents a Smart Agriculture Assistant, an AI-driven system designed to provide farmers with real-time decision support through intelligent data analysis and automation. The system integrates machine learning, and satellite data to monitor soil health, detect crop diseases, and predict weather and market trends. A chatbot-based interface enables farmers to interact with the assistant in their local language, ensuring accessibility and ease of use. The proposed solution aims to enhance crop productivity, reduce losses, and promote precision farming practices. Experimental results demonstrate improved accuracy in disease detection and price forecasting compared to existing approaches. The Smart Agriculture Assistant t

Abstract: The growth of digital technologies has drastically transformed modern life, but it has also given rise to an unprecedented wave of cyber threats. Cyber security is the discipline that protects digital information, systems, and networks from unauthorized access, damage, and disruption. This research paper aims to explore the fundamentals of cyber security, various types of cyber-attacks, security techniques, and future trends. It emphasizes the importance of the CIA triad—Confidentiality, Integrity, and Availability—along with real-world incidents that highlight vulnerabilities in digital infrastructure. The paper also examines tools, methodologies, and advanced technologies such as Artificial Intelligence (AI), Blockchain, and Quantum Computing that can strengthen global cyber defense systems. The goal is to create awareness and outline a path toward a more secure digital future.

Keywords: Cyber Security, Network Security, Cyber Attacks, Encryption, Artificial Intelligence, Blockchain, Data Protection

Abstract: The effectiveness of modern Artificial Intelligence (AI) in personal finance is severely limited by the fragmentation and inaccessibility of an individual's complete financial footprint. Current AI tools can only answer general questions because they lack secure, structured access to real-time, personalized data scattered across multiple platforms (banks, mutual funds, stocks). This paper proposes the design and implementation of a Personal Finance AI Infrastructure leveraging a Decentralized Ledger (Blockchain) model to solve this challenge. The infrastructure creates a single, immutable, and verifiable record of an individual's financial history. By employing cryptographic hashing (SHA-256) and user-controlled smart contracts, the system guarantees privacy, promotes user control, and provides the necessary structured, secure data access required for meaningful, personalized AI financial reasoning, thereby overcoming the limitations of current centralized systems.

Keywords: Blockchain, Decentralized Ledger Technology (DLT), Personal Finance, AI Infrastructure, Data Security, Smart Contracts.

Abstract: Contactless human–computer interaction has become increasingly important for accessibility, hygiene, and natural user experience. This project presents a practical Gesture Controlled Virtual Mouse with Voice Commands framework that allows users to control the computer cursor and perform system actions using only a standard webcam and microphone. The proposed system leverages Google’s Media Pipe for reliable hand landmark detection and OpenCV for image preprocessing and frame handling, while Speech Recognition (with local microphone input) maps spoken commands to system-level actions. The design emphasizes low computing requirements, real-time responsiveness, and user accessibility. Extensive experimental evaluation under multiple lighting and background scenarios demonstrate an average gesture recognition accuracy of ~96%, voice command recognition accuracy of ~93%, an average latency of ~45 ms, and stable operating frame rates (25– 30 fps) on commodity hardware. The system is lightweight, platform flexible, and suitable for applications in healthcare, education, and assistive technologies.

Keywords: Gesture Recognition, Speech Recognition, MediaPipe, OpenCV, PyAutoGUI, Human–Computer Interaction, Virtual Mouse.

Abstract: The Online Traffic Offense System is a web-based platform developed to automate the process of recording, managing, and monitoring traffic violations. The system enables traffic authorities to maintain an online record of offenders, issue fines digitally, and provide transparency in traffic management. Using PHP for the front end and MySQL as the database, this system simplifies data retrieval and minimizes manual errors. This paper discusses the architecture, modules, and implementation of the system designed to enhance efficiency in traffic enforcement.

Keywords: Traffic Management, PHP, MySQL, Offense Automation, Web Application

Abstract: In the modern digital era, social media platforms have become a dominant medium for communication, self-expression, and information exchange. This study explores user behavior, awareness, and perceptions related to sentiment analysis and privacy concerns in social media usage. A structured online survey was conducted to collect responses from participants across different backgrounds, focusing on their level of engagement, familiarity with sentiment analysis, and trust in AI-based emotional interpretation. The analysis reveals that while users are highly active on social platforms and aware of sentiment analysis to some extent, concerns regarding data privacy and ethical implications remain prevalent. Participants expressed mixed views on the accuracy and reliability of AI in detecting emotions, emphasizing the need for transparency and secure data handling. The findings further highlight how sentiment analysis can enhance AI-driven add-on course recommendation systems, enabling more personalized learning based on student sentiments and preferences. The study concludes that balancing innovation with ethical responsibility is essential for the sustainable use of sentiment analysis in both digital and educational environments.

Keywords: Artificial Intelligence, Course Recommendation System, Data Privacy, Sentiment Analysis, Social Media Usage, User Awareness

Abstract: Online job portals are widely used for finding employment, but they are also exploited by scammers who create fake job postings to deceive job seekers. These fraudulent postings often appear legitimate and can lead to identity theft, financial loss, and misuse of personal information. This work proposes a machine learning-based approach to automatically detect fake job posts by analyzing textual and descriptive features from job advertisements. The dataset used in this study is sourced from Kaggle and consists of both real and fake job listings. The text data is preprocessed and transformed into numerical form using TF- IDF, and class imbalance is handled using SMOTE. Several machine learning models including Logistic Regression, Random Forest, and XGBoost were trained and evaluated. Among these, the XGBoost model achieved the highest performance with an accuracy of approximately 97.5%, demonstrating its effectiveness in identifying fraudulent job postings. This system can assist job platforms and users in improving trust and safety by filtering out scam job posts automatically.

Keywords: Fake Job Posts, Machine Learning, XGBoost, TF- IDF, SMOTE, Online Recruitment Fraud Detection.

Abstract: This research focuses on developing a web-based Disease Prediction System using Machine Learning (ML) and the Django framework. The primary objective of the system is to predict possible diseases based on the symptoms entered by the user and to recommend suitable medications and precautionary measures. Machine learning algorithms are trained on a comprehensive medical dataset containing symptoms, diseases, and their interrelationships to ensure accurate predictions. The integration of Django enables a dynamic and interactive web interface that allows users to easily input their symptoms and obtain real-time predictions. The proposed model aims to assist both patients and healthcare professionals by enabling early disease identification, enhancing clinical decision-making, and minimizing the chances of human error in manual diagnosis. Overall, this system provides an intelligent, efficient, and user-friendly approach to disease prediction and preventive healthcare.

Keywords: Disease Prediction, Django Framework, Machine Learning, Healthcare System, Web Application.

Abstract: The advancements in deepfake technology have come swiftly, allowing for the creation of extremely realistic altered images, videos, and audio material. Although there has been considerable progress in unimodal detection in current research, most approaches tend to concentrate on a single modality. This paper analyses more than 20 cutting-edge studies on deepfake detection and pinpoints significant research shortcomings, including the absence of multi-modal frameworks, limitations in datasets, lack of robustness, and insufficient interpretability. To address these issues, we built a prototype detection system based solely on single-modality images that employs two models: a custom Convolutional Neural Network (CNN) and Exception CNN. Our findings underscore the necessity for solutions that incorporate multiple modalities. We suggest an integrated framework for multi-modal detection encompassing images, videos, and audio, which represents the next advancement toward reliable and effective detection systems.

Keywords: Deepfake detection, CNN, Captioned, multi-modal system, Video and Audio Forensics etc.

Abstract: With the rapid digitalization of health monitoring systems, accurate calorie estimation has become a necessity for individuals aiming to manage fitness and weight effectively. Traditional calorie calculators depend on generalized formulas or physical fitness devices, which often fail to capture individual physiological variations such as metabolism, heart rate, and body temperature. This research presents a Machine Learningbased Calorie Burn Tracker that uses six input features — Age, Weight, Gender, Exercise Duration, Body Temperature, and Heart Rate — to predict calorie burn precisely. The system is implemented in Python using the PyCharm IDE, and employs Linear Regression as the core prediction model. Supporting tools such as Django (for web interface development) and Joblib (for model serialization and deployment) enhance usability. A comparative performance study with Decision Tree and Random Forest algorithms confirms that Linear Regression provides optimal results with over 94% accuracy and minimal computation time. The proposed system delivers a cost-effective, scalable, and reliable health tracking solution suitable for integration into mobile and IoT-based platforms.

Keywords: Machine Learning, Calorie Prediction, Linear Regression, Django, Joblib, Python, Fitness Tracking, Health Analytics.

Abstract: As cyber threats grow in complexity and frequency, traditional security systems often fall short in detecting and responding to sophisticated attacks. Artificial Intelligence (AI) has emerged as a transformative force in cyber security, offering advanced capabilities for threat detection, anomaly identification, and real-time response. This paper explores how AI technologies such as machine learning, natural language processing, and neural networks are being integrated into cyber security protocols to enhance their efficiency and adaptability. We discuss current AI-driven applications, including intrusion detection systems, behavioral analysis, and automated threat intelligence. Furthermore, we address challenges such as adversarial AI, ethical concerns, and data privacy implications. The study concludes that while AI significantly strengthens cyber security infrastructures, continuous innovation and regulation are required to manage the evolving threat landscape effectively.

Abstract: In the field of educational data mining, it has become more and more crucial to accurately forecast student performance in order to facilitate early interventions and enhance academic results. In order to predict academic accomplishment, this study uses a dataset of 6000 students (student-scores-6k.csv) that includes factors including study hours, attendance, extracurricular activities, part-time employment, and gender. We used and compared two machine learning algorithms: Random Forest Regressor and Linear Regression. When compared to Linear Regression (R2 = 0.62, RMSE = 8.5), the Random Forest model performed better (R2 = 0.82, RMSE = 5.1). Gender had no bearing on student progress, while weekly self-study hours and absence days were the most significant indicators, according to feature importance analysis. In addition to offering educators and policymakers useful insights for creating interventions that support academic performance, the study shows that non-linear models are more adept at capturing the complexity of educational data.

Keywords: Student Performance, Machine Learning, Regression, Random Forest, Educational Data Mining, Predictive Analytics

Abstract: Deepfake technology has rapidly evolved, enabling the creation of highly realistic synthetic media that can deceive both humans and machines. This research aims to develop a robust deepfake detection framework by integrating deep learning techniques with forensic feature analysis. The proposed system extracts spatial, temporal, and physiological inconsistencies from video and image data to identify synthetic manipulations effectively. Advanced neural architectures such as convolutional neural networks (CNNs) and transformer-based models are combined with handcrafted forensic cues like frequency domain artifacts and texture irregularities. Experimental evaluation on benchmark datasets demonstrates improved accuracy and resilience against adversarial deepfakes. The results indicate that hybrid learning–forensic models offer a promising direction for enhancing media authenticity verification.

Keywords: Deepfake Detection, Forensic Features, Deep Learning, Convolutional Neural Networks (CNN), Transformer Models, Media Forensics, Hybrid Framework, Adversarial Robustness, Synthetic Media, Video Authentication

Abstract: Artificial Intelligence (AI) has emerged as a transformative force across industries, but its extensive use has raised serious environmental concerns. The increasing energy demand, carbon emissions, and e-waste generated by AI systems have significant environmental consequences. This paper explores how the overuse of AI technologies contributes to environmental degradation, focusing on data centers in India and their ecological footprint. It also proposes sustainable strategies for mitigating these effects through green computing, renewable energy adoption, and energy-efficient AI models.

Keywords: Artificial Intelligence, Environmental Impact, Data Centers, Energy Consumption, Sustainability, Green AI.

Abstract: Phishing, a form of cyber-attack in which perpetrators employ fraudulent websites or emails to Deceive individuals into divulging sensitive information such as passwords or financial data, can be mitigated through various machine-learning algorithms for website detection. These algorithms, including decision trees, support vector machines, and Random Forest, analyze multiple website features, such as URL structure, website content, and the presence of specific keywords or patterns, to ascertain the likelihood of a website being a phishing site. This comprehensive review elucidates the concept of phishing website detection and the diverse techniques employed while summarizing previous studies, their outcomes, and their contributions. Overall, machine learning algorithms serve as a potent tool in the identification of phishing websites, thereby safeguarding users against falling prey to such malicious attacks.

Keywords: Phishing Detection, Machine learning, Phish Tank

Abstract: The modern era of digital transformation necessitates the development of highly adaptive and resilient intelligent systems, which has critically highlighted a fundamental paradigm divergence between Human Learning (HL) and Machine Learning (ML). HL is intrinsically rooted in context, abstract reasoning, and ethical frameworks, deriving its power from understanding. Conversely, ML is driven by statistical pattern recognition and computational optimization, relying on optimization. This paper conducts a systematic, interdisciplinary comparison across crucial performance indicators, including data efficiency, generalization capability, common-sense reasoning, and bias vulnerability. The analysis reveals a critical strategic trade-off: ML provides superior speed, scalability, and consistency (low noise), yet it is fundamentally limited by a lack of contextual understanding and a dangerous susceptibility to amplifying systemic algorithmic bias embedded in training data. In stark contrast, HL demonstrates exceptional data efficiency, often exhibiting "less-than-one-shot" learning, coupled with indispensable ethical judgment. The study concludes that the future potential lies in strategic convergence. This is achieved through the development of Hybrid Intelligence systems, facilitated by Neural-Symbolic AI architectures, and governed by robust transparency measures, such as the XAI for Responsible and Ethical AI (XAI4RE) framework, thereby merging human contextual oversight with machine computational precision for trustworthy decision-making.

Keywords: Hybrid Intelligence, Explainable AI (XAI), Common-Sense Reasoning, Data Efficiency, Algorithmic Bias, Neural-Symbolic AI.

Abstract: Water quality assessment and prediction are crucial for environmental management and public health. This research delves into a comprehensive analysis of a water quality dataset, employing standard methodologies for Water Quality Index (WQI) calculation and leveraging advanced machine learning techniques for predictive modeling. The study meticulously details the data loading, preprocessing, WQI computation, and data labeling processes. A comparative analysis of four prominent classification algorithms.Random Forest (0.9718 accuracy), Support Vector Machine (SVM) (0.8250 accuracy), XGBoost (0.9750 accuracy), and Logistic Regression (0.8843 accuracy) is presented, highlighting their performance in classifying water quality into distinct categories. The findings reveal the exceptional predictive capability of the XGBoost model on this dataset, achieving perfect evaluation scores. Visualizations are included to illustrate the distribution of water quality and the comparative performance of the models. This research contributes to the application of machine learning in environmental monitoring and provides a robust framework for predicting water quality

Keywords: Water Quality Prediction, Machine Learning, Random Forest, XGBoost, Support Vector Machine, Logistic Regression.

Abstract: This paper proposes and experimentally validates a production-ready hybrid machine learning framework for intelligent travel destination recommendation, combining collaborative filtering and content-aware modelling within an integrated Flask-based deployment environment. We construct a consolidated dataset of 1,257,000 anonymized user–destination interactions aggregated from multiple open travel datasets and simulated user profiles. The system employs a modular feature-engineering pipeline that extends a 12-dimensional raw feature space (user demographics, ratings, destination tags) into 210 engineered descriptors incorporating semantic embeddings, recency-weighted interaction scores, and geo-temporal correlations.Training was performed with stratified 5-fold cross-validation and temporal validation splits to prevent leakage. On the reserved evaluation set, the proposed model achieved Precision@10 = 0.314, Recall@10 = 0.283, MAP@10 = 0.301, and RMSE = 0.925, outperforming baseline collaborative filtering (Precision@10 = 0.211). Flask-based deployment yielded mean response latency of 78 ms per query under concurrent load, confirming suitability for real-time applications. Ablation studies revealed the largest marginal gain from the semantic-content embedding layer, enhancing personalization for cold-start users. The system requires no proprietary data and is deployable on commodity hardware, providing a reproducible and scalable baseline for academic and industrial tourism analytics. Future directions include integration of context-aware deep models, federated personalization, and reinforcement-based travel itinerary optimization.

Keywords: Hybrid Machine Learning, Collaborative Filtering, Intelligent Travel Recommendation, Matrix Factorization.

Abstract: As education continues to evolve and learners seek personalized experiences, traditional learning systems often struggle to meet these needs. AI is emerging as an intelligent learning assistant that understands, adapts, and teaches in real-time. It takes inputs like a syllabus, subject, prompt, or uploaded notes and automatically creates courses that include videos, text, diagrams, and interactive quizzes. By leveraging machine learning and natural language processing, it can generate question banks, assignments, PYQs, mock tests, and assist with any details required.AI continuously learns from student performance, tracks progress, analyzes areas of weakness, and updates content to provide a personalized and adaptive learning and assessment experience for every learner.

Keywords: AI in Education, Personalized Learning, Adaptive Learning Systems, AI Course Generation, Machine Learning, Natural Language Processing (NLP), Automated Question Bank, Assignment Generation, Interactive Quizzes, Virtual Tutor, Educational Chatbot, PYQ Generation, Mock Tests, Learning Analytics, Student Performance Tracking, Intelligent Assessment, AI-Powered Learning Platform.