VOLUME 14, ISSUE 7, JULY 2025

Abstract: The complexity of public health clinical trials, particularly those orchestrated by federal agencies such as the NIH and CDC demand robust, adaptive, and scalable technologies to ensure timely execution and reliable outcomes. This research investigates the application of Artificial Intelligence (AI) and Machine Learning (ML) techniques for risk forecasting and site performance optimization in multisite clinical trials. The study focuses on predictive models designed to identify operational risks (e.g., patient dropout, protocol deviations, enrollment delays) and evaluates how these models can improve trial logistics, resource allocation, and site selection processes. To enhance real-time monitoring and decision-making, the proposed AI framework is integrated into AWS-native environments using tools like Apache Airflow for pipeline orchestration, EC2 for scalable compute resources, and AWS Lambda for event-driven processing. Through simulation and case study analysis, we demonstrate how the system facilitates adaptive responses to public health emergencies such as pandemics, vaccine trials, or regional disease outbreaks. Furthermore, the study explores practical deployments within the NIH and CDC clinical research ecosystem, illustrating how AI-driven dashboards can aid in forecasting operational bottlenecks, automating compliance reporting, and enhancing site-level performance visibility. The outcomes suggest that AI-integrated platforms not only increase efficiency but also significantly reduce trial risks and costs. The findings support a paradigm shift in how large-scale public health trials are managed, offering a blueprint for future-ready, AI-powered clinical research infrastructure.

Keywords: Public Health Clinical Trials, AI-Enabled Risk Forecasting, Site Perfomance Optimization, Machine Learning in Clinical Research, AWS for Clinical Trials, Apache Airflow, EC2 and AWS Lambda, NIH Emergency Response, CDC Trials, Predictive Analytics, Trial Logistics Optimization, Real-Time Site Monitoring, Federated Trial Intelligence, Adaptive Trial Management

Abstract: In the face of growing water scarcity due to rapid urbanization, population pressure and climate variability, rainwater harvesting (RWH) has emerged as an effective and sustainable strategy for water conservation and management. One of the most critical aspects of designing an efficient RWH system is the accurate estimation of tank size for storage, which must take into account site-specific parameters such as rainfall intensity, runoff potential, tank geometry and user needs. Traditionally, the process of designing such systems requires manual calculations and technical expertise, which may not always be accessible to farmers, local engineers or planners. To address this challenge, a user-friendly Decision Support System (DSS) was developed in this study for the estimation of tank size for rainwater harvesting systems. The DSS, built on an Android platform using the freely available MIT App Inventor, combines user inputs with mathematical models to calculate tank dimensions for various geometries, including circular, square and rectangular shapes. Users can select key parameters such as desired storage volume, pond depth and side slope, while the application performs real-time computations using standard geometric and quadratic equations. The tool simplifies complex calculations into an easy-to-use interface, enabling the estimation of tank dimensions such as top and bottom radius, length and width based on storage volume requirements.

Keywords: Decision support system, rainwater harvesting, tank size, MIT app inventor

Abstract: The development of innovative communication systems for individuals with severe mobility impairments has made significant progress. This wireless system uses head movement detection, powered by ESP-32 accelerometers, to capture subtle gestures and translate them into predefined commands. These commands are transmitted wirelessly to a receiver, activating a speech alert system that allows effective communication. This solution greatly enhances the quality of life for individuals with paralysis by offering a reliable, intuitive, and non-invasive means of interaction. It reduces dependence on caregivers and promotes greater independence. Beyond communication, the system can also be adapted for home automation, making it a versatile tool for daily living. Designed to meet the critical needs of users with conditions such as ALS, spinal cord injuries, or strokes, the system fills a crucial gap in assistive technology. Its user-friendly, cost-effective design empowers users to express themselves, make decisions, and interact meaningfully with others. Ultimately, this system not only meets a practical need but also advances the field of assistive technology by restoring dignity and autonomy to those with disabilities.

Keywords: Assistive Technology, Speech Alert System, Mobility Impairments, Paralysis, Spinal Cord Injury, Non-invasive Interface, Human-Computer Interaction formatting.

Abstract: Surveillance systems play a vital role in ensuring security in both public and private spaces. Traditional CCTV systems lack intelligence and require continuous human monitoring. To overcome these limitations, this paper proposes “Cypher Cam Master”, an intelligent surveillance tool that utilizes Artificial Intelligence (AI) and Machine Learning (ML) for real-time object and motion detection. The system uses computer vision techniques and machine learning algorithms to automatically detect suspicious activities or objects. Our solution is non-intrusive, real-time, and significantly reduces human dependency. The model is trained using a curated dataset and validated with live testing. Accuracy and performance metrics were evaluated using algorithms like Convolutional Neural Networks (CNN) and background subtraction methods.

Keywords: Surveillance, Object Detection, Motion Detection, AI, Machine Learning, CNN, OpenCV, Real-Time Monitoring.

Abstract: Artificial Intelligence (AI) is transforming healthcare by offering solutions for diagnosis, treatment, and patient care through machine learning and other cognitive technologies. AI algorithms analyze vast amounts of data to provide insights for more accurate diagnoses, personalized treatment plans, and efficient healthcare delivery. While AI holds great promise, ethical considerations, data privacy, and the need for collaboration between AI experts and medical professionals are crucial for responsible implementation. AI can analyze medical images (like X-rays and MRIs) and patient data to detect diseases at earlier stages, improving accuracy and speed. AI algorithms can analyze individual patient data to tailor treatment plans, predict patient outcomes, and optimize drug discovery.

Keywords: Surveillance, Object Detection, Motion Detection, AI, Machine Learning, CNN, OpenCV, Real-Time Monitoring.

Abstract: This privacy-conscious wearable computing model gives users contextual, real-time control over their data. In response to the increasing need for guaranteed, personalized health analytic, the suggested approach combines guaranteed multiparty computation (MPC) with multi-key fully homo-morphic encryption (MK-FHE) to alter encrypted processing without jeopardizing data confidentiality. The framework prioritizes edge-level encryption and just-in-time user accept mechanisms, ensuring secure data autonomy at every stage from learning to computation, in contrast to traditional cloud-driven subjects. This method establishes the foundation for user-centered, adaptive, and ethically sound digital health ecosystems by reducing centralized exposure and conforming to legal requirements such as GDPR and HIPAA. Keyword: Edge level encryption, Cloud-driven, Homo-morphic, Multiparty computation

Abstract: This work evaluates textual favorability by means of Natural Language Processing (NLP) analysis of sentiment. Favorability is the degree of like or hate of an individual, policy, or product or service. Text preparation, sentiment scoring, sentiment intensity classification and evaluation via machine training and deep learning models are part of the process. Advanced models such as BERT grasp its context, psychological tone, and minute clues. This approach is valuable in feedback from customers systems, brand monitoring, and political analysis as it enables businesses to make informed judgments grounded on public opinion. In this work, favorability capture in machine learning and lexicon-based approaches is compared. Faster and more comprehensible compared to transformer-based approaches like BERT and RoBERTa, which have greater contextual knowledge for sensitive gestures, sarcasm, and domain-specific emotion, lexicon-based techniques like VADER and TextBlob have These models are built using datasets with annotations of real-world attitudes to better separate obvious favorability from generic positivity. The paper also covers linguistic ambiguity, social media writing noise, and sentiment classification subjectivity. One hybrid solution comprising named entity identification, sentiment ratings, and aspect-based analysis is proposed to overcome these problems. This helps to easily monitor sentiment trends around entities or subjects over time. In political forecasting, customer experience enhancement, and reputation management, sentiment-driven favorability analysis may support strategic decisions.

Keywords: sentiment analysis, favorability, NLP, machine learning, deep learning, BERT, emotion detection, text classification, VADER, public opinion.

Abstract: COVID-19's worldwide spread has put a strain on healthcare systems, especially in recognising it quickly. Despite its time and expense, RT-PCR is the most preferred testing technique. This web-based project classifies chest X-rays as COVID-19, Pneumonia, or Normal using deep learning. Clinicians can check for COVID-19 quicker. The project uses CNN models learned on free medical imaging datasets. This model can discriminate COVID-19 radiography indications among additional lung infections and healthy lungs. Web interfaces simplify chest X-ray submissions. After scanning the image, the backend system instantly categorises correctly. This programme is simple and effective. This helps radiologists, doctors, and researchers make decisions. It cannot substitute for clinical examinations, but it is useful when medical competence is lacking or early detection is important. The programme shows how AI may enhance public health response and digital diagnostics can avoid pandemics.

Keywords: COVID-19, Chest X-ray, Deep Learning, CNN (Convolutional Neural Network), Medical Imaging, Image Classification, Web Application, Healthcare AI, Pneumonia Detection, Diagnostic Tool.

Abstract: The specialty coffee shop market in Toronto has become increasingly competitive, making it essential for business owners to understand the factors that drive customer satisfaction and differentiation. This study aims to identify the main themes expressed in Google Maps reviews of Toronto’s specialty coffee shops over the past year, providing actionable insights for entrepreneurs and industry stakeholders. Over 5000 customer reviews were analyzed using BERTopic (Bidirectional Encoder Representations from Transformers Topic), a state-of-the-art topic modeling approach that leverages contextual language understanding to extract clear and meaningful topics from large volumes of text. The analysis revealed distinct positive themes, such as cozy atmospheres and high-quality coffee, as well as negative aspects like unfriendly service and poor value for money. By correlating these topics with review ratings, the study highlights specific opportunities for improvement and differentiation in the market. These findings offer practical value for business planning, enabling coffee shop owners to make data-driven decisions and enhance customer experiences in a crowded urban landscape. Beyond its local insights, this research introduces a scalable analytical framework that can be applied to market research, business planning, and feasibility studies in diverse sectors, empowering others to extract actionable intelligence from large volumes of unstructured textual data.

Keywords: Topic modeling, BERTopic, Google reviews, Specialty Coffee

Abstract: The growing need for verifiable cryptographic systems in the post-quantum era has spurred interest in alternative methods for analyzing public key cryptography. This study introduces a formal approach for modeling RSA encryption and decryption using deterministic finite automata (DFA) and regular language theory. By abstracting the modular exponentiation process into symbolic transitions, we construct a DFA-based model capable of simulating encryption workflows across varying key sizes and input lengths. The simulation, implemented using Python and JFLAP, demonstrates that RSA operations—typically arithmetic in nature—can be reliably represented and executed through automata. Results show accurate ciphertext generation and high execution efficiency, with computational complexity scaling linearly with input and state size. This formal model not only supports correctness validation but also enables traceability and performance profiling, offering a scalable tool for formal verification and cryptographic analysis. These findings position DFA modeling as a promising foundation for future research in lightweight cryptographic design, post-quantum protocol verification, and symbolic security analysis.

Keywords: Finite Automata, Formal Language, Public Key Cryptography, RSA, Security Analysis.

Abstract: This research investigates the essential elements that drive digital transformation in Ontario’s healthcare system through a systematic analysis of ten peer-reviewed articles from 2018 to 2025. The qualitative study employs the Technology-Organization-Environment (TOE) framework to analyze diverse healthcare settings including community-based clinics, primary care and hospital systems. The researcher conducted targeted searches on Google Scholar and PubMed to select articles that focused on digital health implementation efforts in Ontario or offered transferable relevance. The research identifies recurring patterns in three main areas which include technological aspects like EMR usability, interoperability and organizational aspects including leadership involvement, staff education and environmental factors such as policy consistency and intersectoral teamwork. The review unifies evidence from various real-world settings to help healthcare planners; digital health leaders and policymakers create equitable and scalable digital strategies. The study establishes a thematic framework which will direct upcoming digital health initiatives throughout Ontario's changing healthcare environment.

Keywords: Digital Health Transformation, Ontario Healthcare System, Technology-Organization-Environment (TOE) Framework, Health Care Digital Transformation Success Factors.

Abstract: Electroencephalogram (EEG) is a multi-dimensional time-series brain signal that is highly information packed. While an EEG has high potential to serve in medicine (e.g. disease diagnosis, prognosis, pre-disease risk identification), psycho-physiology (e.g. mood classification, stress monitoring, alertness monitoring, sleep stage monitoring), brain-computer interface application (e.g. thought typing, prosthesis control), and many other areas, the classical design of EEG feature extraction algorithms and EEG classifiers is time-consuming and challenging to fully tap into the vast data embedded in the EEG. Deep learning (or deep neural network) which enables higher hierarchical representation of complex data has been strongly suggested by a wide range of recent research that these deep architectures of artificial neural network generally outperform the classical EEG feature extraction algorithms or classical EEG classifiers. In this paper/ research project, deep neural network architectures have been constructed to perform binary classification on an EEG dataset that was shown by traditional EEG feature extraction methods to have no significant difference between its two data pools (resting EEG recorded before and recorded after listening to music). The Convolutional Neural Network (CNN) model constructed in this project has achieved a validation accuracy of 75±1% using the same EEG dataset. Using the top performing CNN architectures, short duration of relaxing music listening is found to affect the EEG signals generated by the frontal lobe more than the other lobes of the brain; and also to affect the EEG generated by the left cerebral hemisphere more than the right hemisphere.

Keywords: Electroencephalogram (EEG), Deep learning (or deep neural network), Convolutional Neural Network (CNN) model, short duration of relaxing music listening, Activation techniques, epoch, validation accuracy.

Abstract: The Motion Activated Pathway Lighting System is a smart lighting solution designed to enhance energy efficiency and safety in both public and private environments. The system automatically activates pathway lights when motion is detected and turns them off after a predefined time of inactivity. It primarily uses IR (Infrared) or PIR (Passive Infrared) sensors to detect movement and an LDR (Light Dependent Resistor) to ensure operation only in low-light conditions, such as during nighttime. The core control unit of the system is based on the Arduino Mega 2560, which processes sensor data and controls the LED lights via a relay module. This system also includes a dimmer function that allows for a gradual fade-in and fade-out of the lights, improving user comfort and extending the lifespan of the LEDs. The entire setup is powered using a rechargeable lithium battery supported by a step-up voltage converter, making it suitable for remote or outdoor installations. Additional components such as heat sinks and cooling fans ensure the stability of the system. The project is ideal for application in residential complexes, parks, parking areas, industrial zones, and institutional campuses where intelligent lighting control can reduce energy usage and improve user experience.

Abstract: Sign language is a good visual communication aid for those with auditory disabilities. This language is also prevalent for those with speech impairment. However, the general populous have little knowledge on sign language, and often find difficulty communicating with someone who is primarily versed in sign language. Our goal is to build a system that can provide robust hand sign-language gesture recognition for SL. This can be of extensive help in public places, especially sign language that isn't often universally understood by the majority of people. This chose four signs used worldwide and prepared a data set for these signs and performed the necessary processing for them, then we chose the SVM algorithm to classify these data, and the algorithm showed a high classification efficiency that reached 99%.

Abstract: Image processing plays a vital role in various techno- logical advancements such as medical imaging, surveillance, au- tonomous systems, and augmented reality. Leveraging techniques such as filtering, enhancement, segmentation, and classification, image processing enables accurate and automated interpretation of visual data. This paper explores the core methodologies, real- world applications, and future directions of image processing. The integration of AI and deep learning has further improved the efficiency of image processing, enabling innovative use cases and expanding its potential. Index Terms: Image Processing, Deep Learning, Computer Vision, Feature Extraction, Classification.

Abstract: Nowadays, choosing a specialization is a crucial decision for students, especially in determining their career path. Decision-making is an essential life skill that involves gathering information, generating alternatives, evaluating options, and reaching a decision. This research design for students at the final stage of secondary school to assist them in choosing their desired discipline of study. The research will use the Support Vector Machine (SVM) to examine the criteria for selecting university specialties that are best suited for each student's needs, interests, and strengths. It will also provide a chance for students to explore potential outcomes and opportunities associated with those specialties, helping them make informed decisions that will shape their future. Understanding the different steps in the decision-making process will help students make informed choices that will shape their future.

Keywords: Specialization, Students, Decision, University, Future, SVM.

Abstract: Stem cell research is pivotal in advancing regenerative medicine and biological studies. However, accurately identifying stem cells amidst a heterogeneous population of cells remains a significant challenge, often requiring labor-intensive manual processor advanced equipment. This project proposes an AI-powered device for accurate stem cell detection, combining deep learning techniques with advanced image analysis to automate and enhance the identification process. The device employs a convolutional neural network (CNN) trained on labeled microscopic images of stem cells, enabling precise classification based on unique cellular features. The model is integrated into a user-friendly software system, capable of analyzing static images or real-time video feeds, providing instant and reliable results. Key innovations include robust data pre-processing using augmentation techniques to improve model generalization, real-time detection capabilities, and adaptability for diverse imaging setups. By leveraging the power of AI, this solution reduces the need for extensive manual effort, minimizes error rates, and accelerates the work flow in laboratory and clinical settings. This project demonstrates the potential of artificial intelligence in biomedical applications, aiming to democratize access to efficient diagnostic tools and stream line stem cell research processes. Future enhancements include hardware integration for portable use and application expansion to other cell types and biomedical imaging challenges

Keywords: convolutional neural network, diverse image, clinical setting.

Abstract: In the current digital era, financial cybersecurity is crucial, where the financial industry is vital to the world economy. The frequency and sophistication of cyber-attacks are mounting, making difficult for traditional fraud detection systems to stay up with the changing threats. As a consequence, using Artificial Intelligence (AI) based Machine Learning (ML) approach in fraud detection system, the presented article offers an enhancing financial cybersecurity. The collected input data from financial and transactional data gets pre-processed with the help of data cleaning, data normalization, which aims to remove noise and improve the quality of input data. For effectual forecasting of fraud prediction, the proposed model uses a novel hybrid rule based and isolation forest approach. This rule based scheme ensures regulatory compliance and interpretable alerts, while the isolation forest proficiently isolates anomalies without requiring labeled data. Overall, the analytical evaluation on real world financial transactions system is ensured by the introduced topology, which accomplishes lower errors and higher accuracy of (97.45%) with a significant reduction in false positives and faster decision making compared to the traditional supervised learning models.

Keywords: Financial cybersecurity, cyber-attacks, fraud detection systems, Artificial Intelligence, Machine Learning, Hybrid rule based and isolation forest.

Abstract: Agriculture is of utmost importance to the Indian economy. The production of main crops such as rice, maize, tomatoes, and potatoes go a long way to affect the livelihoods of the farmers. However, these crops are highly susceptible to many challenges most especially diseases that attack them; such maladies drastically reduce productivity. Early and rapid identification of such diseases are critical for initiating appropriate measures to contain potential losses. Deep learning techniques will be harnessed in this study involving feature extraction from digitized images of diseased plants for the accurate identification of maladies. Deep learning has also previously proven an efficient tool in handling very large datasets and finding patterns between normal and anomalous leaves. This review looks at different deep learning algorithms like VGG16, VGG19, RegNet50, EfficientNet etc. used in different studies and checks the accuracy, efficiency, and reliability of these models in detecting diseases in crops. The information learned from this review will help to find out the best deep learning algorithms for crop diseases detection. By better identifying and handling diseases, this study aims to increase productive crop farming in India which will help the sustainable growth of the agricultural sector.

Keywords: Crop Disease Detection, Convolutional Neural Networks, Image Classification, Deep Learning, Transfer Learning, Internet of Things.

Abstract: Virtual Reality (VR) is revolutionizing the educational sector by offering immersive, interactive, and engaging environments for learners. This paper presents the development of a cost-effective VR-based learning platform specifically designed for anatomy education. The platform, developed using Unity, allows students to explore human body structures in a 3D virtual environment, enabling better understanding and retention. By reducing dependency on expensive physical models or labs, this system provides an affordable and scalable educational tool. Performance evaluation and user feedback indicate improved learner engagement and comprehension.

Keywords: Immersive learning, VR platform, Cost effective VR, VR for education

Abstract: Loan approval is a critical process in the banking and financial sector, requiring accurate and timely decision-making to ensure effective risk management for institutions and financial support for applicants. Traditional loan processing methods are often manual, time-consuming, and susceptible to human bias or inconsistency, which can result in delayed or inaccurate decisions. To address these challenges, this project proposes a machine learning-based solution using Random Forest, Support Vector Machine, Logistic Regression and Decision Tree algorithms to predict the likelihood of loan approval. The system is trained on historical loan data, including features such as income, employment status, credit history, education level, marital status, and loan amount, to identify meaningful patterns that distinguish approved from rejected applications. Logistic Regression offers a simple and interpretable model for binary classification, while and robustness by aggregating predictions from multiple decision trees. In addition to prediction functionality, the application includes a secure login and registration module, where user credentials are stored in a database to maintain account integrity. Users can enter loan application details through a clean and user-friendly web interface, with all input data securely saved for processing and analysis. The system delivers real-time prediction results, helping applicants quickly understand their chances of loan approval. This intelligent and scalable solution not only reduces the workload on financial officers but also enhances consistency, transparency, and efficiency in the loan approval process, paving the way for smarter decision-making in modern banking systems.

Keywords: Loan Approval Prediction, Machine Learning, Logistic Regression, Random Forest Classifier, Deep Learning, Loan Application System.

Abstract: The greatest technique to track the functionality and health of the cardiovascular system and spot diseases associated with it is to use ECG signals. The ECG pattern reflects irregular heartbeats, and these abnormal signals are referred to as ARRHYTHMIAS. The need of the hour is growing for automated ECG arrhythmia signal categorization and identification that delivers faster and more precise results .Different machine learning techniques have been used to improve the models speed and durability as well as the accuracy of the findings. The architectures and datasets used have received a lot of attention, but preparing the data is also crucial. In this study, a pre-processing method that greatly increases the ECG classification accuracy of deep learning models is proposed alone with a modified deep learning architecture that increases training stability. The system can achieve accuracy levels of more than 99% with this pre-processing method and deep learning model without over fitting the model.

Keywords: Electrocardiogram (ECG), Deep learning (or deep neural network), Convolutional Neural Network (CNN) model, ARRHYTHMIAS, Activation techniques, epoch, validation accuracy.

Abstract: This paper presents a comprehensive full stack Employee Management System (EMS) that combines a modern React frontend with a robust Spring Boot backend. The system aims to streamline HR processes, enabling efficient employee and department management with features like authentication, CRUD operations, dashboards, and data visualizations. It supports MySQL and MongoDB databases, employs REST APIs for data exchange, and is containerized using Docker and Kubernetes for deployment. This research reviews existing literature, compares different technologies, and presents our architecture, implementation strategy, sample circuit diagrams, and source code to highlight the design choices and innovations.

Abstract: The growing sophistication of deepfake technology has created serious challenges for digital forensics, particularly within law enforcement. Deepfakes highly convincing but entirely fabricated audio, video, and image content—pose significant threats to public trust, security, and the integrity of investigations. To counter these risks, this project proposes the development of a unified software solution designed to detect deepfakes across multiple media formats, tailored specifically for cyber police use. The system will combine Machine Learning, Artificial Intelligence, and forensic analysis techniques to uncover tampering and manipulation in digital content. With a multi-layered detection framework, it will analyze visual anomalies and audio inconsistencies, employing Deep Learning models such as CNNs for image and video analysis and RNNs for audio detection to distinguish between genuine and fake media. Trained on extensive datasets, the system enhances detection accuracy and strengthens the fight against digital deception. Furthermore, it will seamlessly integrate with existing forensic tools, empowering investigators to quickly assess the authenticity of digital evidence. This advanced detection platform is intended to aid law enforcement in preventing and investigating crimes involving fraud, identity theft, blackmail, and the spread of misinformation enabled by deepfake technology.

Keywords: Deepfake Detection, Digital Forensics, Convolutional Neural Networks (CNNs), Reccurent Neural Networks (RNNs)

Abstract: The Personality Prediction System is a web-based recruitment platform that leverages machine learning to enhance the hiring process by predicting candidate personalities and recommending suitable job roles. Developed using ASP.NET for the frontend, the system incorporates a Gradient Boosting Machine (GBM) model to analyze responses from a multiple-choice personality test and generate predictions. The platform is structured into three main modules: Admin, Company, and Candidate. Admins can manage and communicate with both companies and candidates, view test results, and oversee system activity. Companies can create and manage job vacancies, initiate personality assessments for applicants, and view the resulting predictions to aid in selection. Candidates can update their profiles, generate resumes using ASP.NET Web API and iTextSharp, take personality tests when requested by companies, and receive tailored job recommendations based on their predicted personality type. The system includes a messaging feature across all modules for effective communication, making it a comprehensive, intelligent, and user-friendly solution for modern recruitment.

Keywords: Personality Prediction, Recruitment System, Machine Learning, Gradient Boosting Machine, Job Recommendation

Abstract: A Weather Forecast & Prediction System is envisioned to leverage the power of Machine Learning (ML) to provide accurate and accessible weather information, as depicted in the provided image. This system would process user input, specifically a city name, to fetch and utilize historical meteorological data encompassing parameters such as temperature, humidity, wind speed, and atmospheric pressure. At its core, the system would employ a suite of well-established ML algorithms for prediction. Linear Regression would be applied for forecasting continuous numerical values like temperature, humidity, wind speed, and pressure, learning the linear relationships between various features and these target variables. For predicting categorical outputs, such as the "Condition" (e.g., Sunny, Rainy, Cloudy), the K-Nearest Neighbours (KNN) algorithm would classify the current or future weather state based on the similarity to historical weather patterns. Furthermore, Random Forest, an ensemble learning method, would be utilized for its robustness and ability to handle both regression and classification tasks, capturing more complex, non-linear interactions within the weather data and providing highly accurate predictions for all mentioned parameters. The system's architecture would involve efficient data acquisition from historical archives and real-time APIs, robust data pre-processing and feature engineering to prepare the data for the ML models, and a user-friendly interface to display the predicted current conditions and future forecasts clearly and concisely.

Keywords: Weather Prediction, Forecasting, K-Nearest Neighbours (KNN), Linear Regression, Random Forest, User Interface (UI), Web Application , and Classification.

Abstract: Massive Open Online Courses (MOOCs) have revolutionized access to education, but their scalability presents significant challenges in effective resource allocation and scheduling. This paper explores the application of Operations Research (OR) techniques to optimize key constraints in MOOC environments—specifically, instructor time management, student learning progress, and grading workloads. By formulating the problem as a multi-objective optimization model, we demonstrate how linear programming (LP), integer programming (IP), and queuing theory can support intelligent resource distribution. Case studies and simulation models show that the application of OR methods can significantly reduce bottlenecks in instructional support, balance grading demands, and enhance personalized student pacing. The findings suggest a hybrid OR framework can be embedded within MOOC platforms to improve efficiency and learning outcomes.

Keywords: Resource Allocation, Scheduling Optimization, Grading Workload, Multi-objective Optimization, Linear Programming.

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 an academic system which 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: This paper presents a survey of AI-powered mental health companions that are intended to enhance mental health and offer emotional support is presented in this paper. These systems simulate conversations and help users with stress, anxiety, and depression by utilising artificial intelligence, which includes chatbots, natural language processing, and emotion recognition. We examine current advancements, widely used applications, important technologies, and difficulties like accuracy, privacy, and moral dilemmas. The future potential of incorporating AI technologies into easily accessible, secure, and encouraging mental health services is also highlighted in the paper. Index Terms—component, formatting, style, styling, insert

Abstract: The "Novel RFID Cloud Based Smart Attendance System" project addresses the persistent challenges faced by organisations and institutions in efficiently managing attendance records for employees, students, or visitors. Traditional attendance methods, often paper-based or relying on manual data entry, are highly susceptible to errors, time-consuming, and crucially, lack real-time monitoring capabilities. This innovative project, developed by Masud Alam, Rahul Singh, Shahbaz Khan, and Krishn Kant Sah from the School of Information Technology at SAM Global University for the 2024-25 academic session, offers a robust solution to these inherent limitations. At its core, the system harnesses the power of Radio Frequency Identification (RFID) technology integrated with a scalable cloud computing architecture. This synergistic approach enables attendance data to be securely stored in a remote database, ensuring it is accessible from anywhere in the world and at any time. The project meticulously combines a range of hardware and software components to achieve its comprehensive functionality. Key hardware elements include the ESP32-WROOM-32E Microcontroller Unit (MCU), which acts as the central control unit, orchestrating communication with peripherals such as the RFID MFRC522 reader module and the OLED display module SSD1306, and managing the solenoid lock. The RFID MFRC522 reader is responsible for automated identification by scanning and reading digitally encoded data from 1k MIFARE Classic Contactless Smart Cards. A 1 Channel 5v Relay Module facilitates the control of the 12V solenoid lock, which grants temporary access for approximately 10 seconds upon a successful RFID scan. The OLED display provides immediate, real-time feedback on attendance status, including student login and logout events, chosen for its clear view angle and pixel density. From a software perspective, the system incorporates a comprehensive web-based user interface (Web UI), meticulously crafted using HTML, CSS, and JavaScript (with the Bootstrap framework). This UI empowers administrators to manage attendance records, view reports, and configure system settings, including the ability to add, edit, and delete user profiles. The backend, primarily developed with PHP, is responsible for fetching and rendering data from the MySQL database, a Relational Database Management System (RDBMS) that handles structured user data storage, retrieval, and management. The Arduino language is utilised to program the ESP32 microcontroller. For deployment, the project leverages AWS Cloud services, specifically EC2 instances for hosting the website on an Apache server within a Linux virtual machine, and AWS S3 for static file storage, with CloudFront providing efficient content delivery globally. The development process adhered to a structured methodology encompassing requirement analysis, detailed system design, hardware setup, software development, rigorous integration and testing, strategic deployment, and ongoing maintenance and support. The system's functional capabilities include automated attendance tracking via RFID scanning, real-time attendance information display on the OLED, and secure access control through the solenoid lock based on attendance status. The Web UI allows for the generation of attendance reports in various file formats based on specified criteria. Robust cybersecurity measures, including data encryption for user passwords and personal details, are embedded within the backend and database to prevent unauthorised access and mitigate risks such as RFID cloning or spoofing. The system's non-functional attributes highlight its user-friendly interface, low latency and quick response times, scalability to accommodate an increasing number of users and hardware components, stability for consistent operation, and minimal downtime. Furthermore, its design and documentation facilitate easy maintenance, troubleshooting, and future enhancements. Having achieved significant milestones, the project has resulted in a functional smart attendance system that effectively demonstrates its core features and meets initial requirements. The market outlook for such an automated attendance solution is promising, driven by the escalating demand for digitisation and accurate tracking across diverse sectors. The project's unique selling proposition lies in its comprehensive integration of hardware, software, and user interface. Target clients encompass educational institutions, corporate offices, government organisations, event management companies, and healthcare centres. Future work will focus on implementing enhanced security features (e.g., biometric authentication), developing advanced reporting and analytics, building a mobile application, and optimising for scalability and performance. Further integrations with Student Information Systems (SIS) and Learning Management Systems (LMS) are also envisioned to streamline data management and synchronisation. Key recommendations include conducting extensive user testing, continuous refinement and documentation, fostering partnerships and collaborations, and developing a strategic marketing plan to ensure the solution's growth and success. .

Keywords: RFID, Cloud Computing, Smart Attendance System, ESP32 Microcontroller, Web User Interface (Web UI), Access Control, Data Management (MySQL), Real-time Display (OLED), Hardware Integration, Security.

Abstract: The precise forecasting of stock market prices presents a tough challenge, owing to the crucial volatility and details of the market. In this research paper, we tackle this challenge by introducing a stock market prediction model grounded in the Random Forest algorithm. Our study centers on historical trading data encompassing a diverse array of stocks and ETF funds, harnessing the capabilities of AI technology and machine learning methodologies to forecast and scrutinize stock prices through regression analysis. The outcomes underscore the Random Forest model's capacity to achieve commendable accuracy in stock prediction, thereby offering invaluable insights for both institutional and individual stock investments. These models rely on technical indicators as inputs, with the closing value of stock prices serving as the predicted variable. The results not only underscore the effectiveness of our proposed approach in constructing predictive models for stock price projection but also highlight the potential of Machine Learning algorithms to reveal valuable insights into the dynamics of stock market activity. Moreover, our paper investigates the exploration of diverse Machine Learning models, encompassing Linear Regression, Support Vector Regression, Decision Tree, Random Forest Regressor, and Extra Tree Regressor. Their implementation has proven instrumental in achieving precision in stock price prediction and has furnished fresh perspectives into the intricate interplay between buyers and sellers in the stock market. The evaluation of these models is grounded in their accuracy in predicting stock prices, using both closing values and stock prices as crucial metrics.

Keywords: stock, random forest, prediction, tree, regressor, Machine Learning

Abstract: In the age of customized medicine, getting the right drug to the right patient at the right time is a major challenge. This research describes the creation of an Adaptive medicine Recommendation System that employs Reinforcement Learning (RL) to improve medicine prescription accuracy based on unique patient characteristics. Unlike standard drug recommendation systems, which rely on static rules or supervised learning models, the proposed system represents the treatment process as a Markov Decision Process (MDP) and using RL approaches to learn optimal drug strategies over time. The system generates dynamic and tailored drug recommendations based on patient-specific data such as age, medical history, present symptoms, and ongoing drugs. It is constantly learning and adapting based on feedback from patient results, making it resistant to changing health circumstances. The RL agent is trained and tested using benchmark healthcare datasets, and its performance is compared to traditional methods in terms of accuracy, flexibility, and safety. The findings show that the suggested approach improves clinical decision-making while also paving the way for intelligent, real-time, and patient-centric healthcare solutions.

Keywords: Reinforcement learning, Markov Decision Process, Tailored drug, centric healthcare solutions.

Abstract: The thyroid gland, one of the body’s key endocrine organs, produces two essential hormones that regulate metabolic activity. Abnormal functioning of this gland can lead to disorders such as hypothyroidism and hyperthyroidism, both of which significantly disrupt the body's normal physiological processes. Although thyroid disorders are generally diagnosed through blood tests, these tests often yield ambiguous or noisy results, making accurate diagnosis difficult. To address this challenge, the present study incorporates data cleaning methods and machine learning techniques to enhance the accuracy of thyroid disease detection and prediction. Clean and structured data improved the reliability of the analysis. Various machine learning algorithms, including logistic regression, decision trees, k-nearest neighbors (KNN), support vector machines (SVM), XG Boost, and artificial neural networks (ANN), were employed to model and predict.

Keywords: hypothyroidism, hyperthyroidism, diagnosed, machine learning algorithms decision tree, KNN, SVM, ANN

Abstract: The outbreak of COVID-19 marked a significant shift in the way work is structured, disrupting traditional job models and accelerating progress in areas such as urban planning, digital connectivity, and corporate operations. This study offers an in-depth analysis of the global impact of remote work on socio-economic systems. By drawing on cross-sectional data, theoretical perspectives, and recent scholarly findings, it explores evolving labour trends, productivity shifts, workforce expectations, ecological effects, and the spatial organization of economic activity. The results provide valuable guidance for leaders in government, industry, and academia. The shift toward remote work has brought about profound changes in the nature of the modern workplace. This research explores how this shift has influenced employment structures, daily routines, productivity levels, and urban development. A key part of the discussion involves comparing remote work with traditional in-office roles. While working remotely allows for increased independence and schedule flexibility, it can also lead to challenges such as limited social interaction, unequal access to digital tools, and potential obstacles to career advancement. On the other hand, working on-site promotes face-to-face collaboration and structured workflows but may fall short in offering the flexibility that many workers now seek.

Keywords: Remote Work, Hybrid Work, Productivity, Labor Economics, Urban Shift, Telecommuting, Work from Home

Abstract: Recent years have seen a significant and widespread rise in false news, which is defined as material that is shared with the intention of defrauding people.This kind of misinformation is dangerous to social cohesion and wellbeing because it exacerbates political polarisation and public mistrust of authority figures.As a result, fake news is a phenomena that significantly affects our social lives, especially in politics.In order to address this issue, this study suggests brand-new methods based on machine learning (ML) and deep learning (DL) for the fake news identification system.This paper's primary goal is to identify the best model that produces high accuracy performance.Hence, in order to identify fake news, we provide an improved Convolutional Neural Network model (OPCNN-FAKE).Using four benchmark datasets for fake news, we assess how well OPCNN-FAKE performs in comparison to Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and The Six Regular ML Techniques: Decision Tree (DT), logistic Regression (LR), K Nearest Neighbor (KNN), Random Forest (RF), Support Vector Machine (SVM), and Naive Bayes (NB). The parameters of ML and DL have each been optimised using the grid search and hyperopt optimization approaches, respectively. Moreover, Glove word embedding has been utilised to encode features as a feature matrix for DL models while N-gram and Term FrequencyInverse Document Frequency (TF-IDF) have been used to extract features from the benchmark datasets for regular ML. Accuracy, precision, recall, and F1- measure were used to validate the data in order to assess the performance of the OPCNN-FAKE. Compared to other models, the OPCNN-FAKE model has the best performance for each dataset.

Keywords: Fake News Detection, OPCNN-FAKE, Deep Learning, Natural Language Processing, BERT, Misinformation, Text Classification

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: The proliferation of Artificial Intelligence (AI) in education presents immense opportunities, yet it risks widening the digital divide, particularly for students in rural areas or those with limited technical exposure. Many students find it challenging to effectively leverage complex AI tools or navigate vast online resources, especially under the pressure of last-minute exam preparation. The present study introduces "Last Leap," an integrated web-based AI study assistant designed to democratize access to AI-powered learning aids. Last Leap provides a user-friendly interface for students to instantly generate comprehensive notes, concise summaries, relevant video recommendations, and practice with interactive quizzes and flashcards directly from a user-inputted topic. Furthermore, a context-aware AI chatbot offers immediate clarification on the generated material, and all generated content can be saved to personalized accounts and exported as formatted PDFs for offline study. The platform aims to simplify AI interaction, reduce information overload, and enhance study efficiency, making advanced learning tools more accessible and impactful for a broader student population. Preliminary findings suggest a significant reduction in time spent on note generation and an increase in student engagement with targeted learning materials.

Keywords: Artificial Intelligence in Education, Digital Divide, E-Learning, Study Assistant, Note Generation, Automated Quizzing, User Experience.

Abstract: In the era of smart automation, AI-powered voice assistants have evolved beyond basic task execution to provide intelligent decision-making capabilities that significantly enhance human-computer interactions. With the integration of advanced machine learning algorithms and real-time data processing, modern voice assistants are not just reactive tools but proactive systems capable of learning and adapting to user behavior over time. This paper presents "Neon AI," a next-generation, customizable AI voice assistant developed using Python, designed to bridge the gap between standard virtual assistants and the growing need for personalized, context-aware automation. Neon AI incorporates robust voice recognition techniques, state-of-the-art natural language processing (NLP), and intelligent task automation capabilities to streamline user interactions across various platforms. Leveraging powerful AI models like GROQ and Cohere, Neon AI offers dynamic and adaptive responses tailored to individual preferences, making it versatile for both personal and professional applications. Additionally, the assistant features a user-friendly graphical interface developed using PyQt5, enhancing accessibility for users with varying technical backgrounds. The paper highlights the comprehensive methodology adopted in designing the modular architecture of Neon AI, details the implementation processes, and provides an in-depth analysis of the system's performance through empirical metrics and graphical evaluations. Experimental results demonstrate Neon AI's proficiency in handling multi-faceted tasks, delivering high accuracy in voice recognition, low response latency, and an engaging user experience, paving the way for future enhancements in AI-driven personal assistant technologies.

Keywords: AI Voice Assistant, Task Automation, Natural Language Processing, Neon AI, Voice Recognition, GROQ API, Speech Recognition, PyQt5, Automation

Abstract: Unauthorized two-wheeler access through pedestrian pathways at significant safety and regulatory concerns, leading to traffic violations and pedestrian inconvenience. Traditional systems rely on manual monitoring or basic sensors, which are inefficient and prone to errors. To address this issue, our project implements an AI-powered automated gate control system using ESP8266 as the main controller. An camera captures real-time video and pass on it to a server hosting a YOLOv8-based AI model, which accurately differentiates between motorcycles and humans. If a pedestrian is detected, the gate remains open, ensuring smooth passage, whereas if a motorcycle is identified, the wifi-contoller triggers the gate to close, preventing motorcycle access. In this system an ultrasonic sensor is used to measure the distance between the detected entity and the gate. This automated approach put an end to the need for manual monitoring, ensuring a safe, efficient, and intelligent pathways system that enhances pedestrian safety and enforces traffic rules. By merging AI-driven image processing with IoT-based hardware control, this system effectively prevents motorcycle entry while allowing pedestrian movement, thereby improving overall pathways management and security.

Keywords: Pedestrian Safety, Urban Environments, IoT-Based System, Ultrasonic Sensors, Infrared Sensors, Real-Time Alerts, Sustainable Ecosystem.

Abstract: Tomato cultivation is susceptible to various diseases, leading to significant yield loss and economic impact. Rapid and accurate prediction is essential for timely intervention and mitigation. Deep learning techniques, specifically Convolutional Neural Networks (CNN), are applied for automated detection of tomato leaf diseases. The methodology involves acquiring high-resolution images of tomato leaves and training a CNN model to classify them into healthy or diseased categories. The dataset comprises labeled images representing Early Blight, Late Blight, and healthy leaves. The CNN architecture is optimized to achieve high accuracy, precision, recall, and F1-score. The trained model demonstrates promising results in identifying tomato leaf diseases even under environmental variations and leaf deformities. The approach also allows for near real-time detection, enabling timely agricultural interventions. This research contributes to automated agricultural monitoring systems, aiding farmers in early disease detection and management, thereby enhancing crop productivity and sustainability.

Keywords: Tomato Leaf Disease Detection, Convolutional Neural Network (CNN), Deep Learning, Image Classification, Early Blight, Late Blight, Real-time Detection, Precision Agriculture

Abstract: Brain tumours are potentially fatal anomalies in neural tissues that need to be identified quickly and properly classified in order to be treated. This study offers a deep learning and image processing framework for automated brain tumour detection. The study combines preprocessing, feature extraction, and classification into a single model by using convolutional neural networks (CNNs) to recognise and categorise different types of tumours from MRI scans. When compared to traditional machine learning techniques, experimental results show notable increases in accuracy. The suggested approach provides a dependable tool to help radiologists make clinical decisions more quickly, reduce human error, and aid in early diagnosis.

Keywords: Brain tumour, MRI, Deep Learning, CNN, Image Processing, Medical Diagnosis

Abstract: Prion diseases are rare but invariably fatal disorders affecting the nervous system [8]. Identifying them at an early stage is complicated when working with large-scale omics data, as the datasets often contain few patient samples and many irrelevant or overlapping features [9]. In this work, we employ a genetic algorithm (GA) to perform feature selection, integrated with a Random Forest (RF) classifier for prediction [10]. Experiments on synthetic biomarker datasets, followed by external testing, showed that the GA could isolate concise feature sets that enhanced model generalization [11]. The final configuration reached a hold-out accuracy of at least 0.97 and achieved 0.94 accuracy on an unseen test set [12]. We detail the methodology, performance trends, selected features, and the potential impact on biomarker identification and early clinical diagnostics.[13]

Keywords: Prion Disease; Transmissible Spongiform Encephalopathy; Genetic Algorithm; Feature Selection; Random Forest; Biomarkers [14].