Journal of Science Engineering Technology and Management Science

Automated video summarization remains a central and unresolved challenge in multimedia understanding: the exponential growth of video content across streaming platforms, surveillance systems, and educational repositories demands efficient mechanisms that can distill hours of footage into concise, semantically coherent representations without human intervention. This paper presents the Adaptive Video Summarization framework (AVSUM), a unified multi-modal architecture that automates keyframe selection and temporal segment extraction for arbitrarylength video inputs. AVSUM integrates three tightly coupled modules: (i) a Multi-Modal Feature Extraction Engine that processes both visual frames and audio-derived transcripts, producing a joint spatio-temporal embedding termed the Video Semantic Signature (VSS); (ii) a Temporal Attention Selector, a trained sequence model that maps VSS embeddings to importance scores and produces ranked keyframe sequences supported by SHAP-based saliency explanations; and (iii) a Scene Coherence Advisor that diagnoses temporal redundancy and prescribes adaptive compression strategies. Unlike existing summarization pipelines that operate as opaque extraction loops, AVSUM delivers interpretable, evidence-backed segment rankings through a self-improving Knowledge Base seeded from benchmark video corpora. Evaluated across 60 benchmark video datasets spanning lecture recordings, news broadcasts, surveillance footage, and cinematic content, AVSUM achieves an F1-Score@Summary of 0.74, a 42-percentage-point improvement over random keyframe selection and a 19-percentage-point improvement over the uniform-sampling heuristic (both significant at p < 0.001, Wilcoxon signed-rank test). The system is deployed as a full-stack interactive web application built with Python, PyTorch, OpenCV, FastAPI, and React. AVSUM makes six original contributions to the video summarization literature, including the Video Semantic Signature embedding scheme, multi-modal redundancy analysis, and a closed-loop self-improvement mechanism— capabilities absent from existing open-source tooling.

Keywords : video summarization, keyframe extraction, temporal attention, multi-modal learning, scene segmentation, explainable AI, SHAP, OpenCV, PyTorch, Video Semantic Signature

Author : Sumit Kumar, Mohapatra Girashree Sahu, Debadatta Sahoo, Satya Ranjan Pattanaik

Title : AVSUM: Adaptive Video Summarization Using Multi-Modal Temporal Attention and Scene-Aware Keyframe Extraction

Volume/Issue : 2026;03(06)

Page No : 223-233

The Transport Management System (TMS) is a fullstack web-based application developed to simplify and automate transportation management activities digitally. Traditional transport management methods involve manual booking, vehicle allocation, route management, and driver coordination, which consume significant time and effort. The proposed system provides a secure, efficient, and user-friendly platform for managing transportation services online. The system is developed using the MERN Stack, including MongoDB, Express.js, React.js, and Node.js. It contains two separate interfaces: a React frontend for users and an Expressbased admin panel for administrators. Users can register, log in securely, browse vehicles, drivers, and routes, create bookings, and monitor booking status. Administrators can manage drivers, vehicles, routes, bookings, and assign vehicles or drivers through a dedicated dashboard. The system supports JWT authentication, REST APIs, centralized database management, automated booking workflows, and responsive user interfaces. The project improves operational efficiency, reduces paperwork, enhances booking management accuracy, and provides better scalability for modern transportation services.

Keywords : Transport Management System, MERN Stack, React.js, Node.js, MongoDB, Express.js, JWT Authentication, Booking Management, Vehicle Management, Driver Management, REST API, Web Application, Admin Dashboard, Route Management, Full Stack Development.

Author : Rajendra Behera, Priyabrata Nayak

Title : Transport Management System

Volume/Issue : 2026;03(06)

Page No : 234-239

The rapid growth of online food delivery services and cloud-based web applications has significantly increased the demand for scalable, secure, and efficient food delivery management systems. Traditional food ordering systems often suffer from delayed communication, limited delivery tracking, inefficient workflow coordination, and lack of real-time operational visibility. This paper presents the design and implementation of “Pronto,” a MERN-stack-based Food Delivery Platform aimed at improving food ordering, delivery coordination, payment management, and real-time tracking in modern digital commerce environments. The proposed system integrates real-time geolocation tracking, secure payment processing, cloud-based media management, and multi-role workflow coordination to enhance operational efficiency and customer satisfaction. The platform provides live delivery tracking through GeoJSON-based location handling and reverse geotagging mechanisms, enabling customers to monitor delivery progress dynamically. The system also incorporates Firebase Authentication, Razorpay payment integration, Cloudinary cloud storage, Redux Toolkit state management, and responsive dashboard visualization to ensure secure communication, scalability, and improved user experience. The application is developed using React.js, Tailwind CSS, Node.js, Express.js, and MongoDB, ensuring modular architecture, responsive frontend rendering, and secure REST API communication. The implementation results demonstrate improved delivery transparency, secure transaction handling, workflow coordination, and operational efficiency compared to traditional food delivery systems..

Keywords :

Author : Mr. Suvamjit Mishra, Mr. Swadhin Jena, Prof. Smruti Smaraki Sarangi

Title : Design and Implementation of an MERN-Based Food Delivery App with Live Tracking

Volume/Issue : 2026;03(06)

Page No : 240-249

The Online Exam Management System is a webbased application developed to automate and simplify the process of conducting examinations in educational institutions. Traditional examination systems mainly depend on paper-based processes, manual evaluation, physical invigilation, and handwritten record management, which consume significant time and are prone to human errors. The increasing demand for digital education and remote learning environments has created the need for secure,scalable, and automated online examination platforms. The proposed Online Exam Management System provides a centralized platform where administrators can create examinations, manage questions, monitor students, and evaluate results efficiently. Students can register, log in securely, attend online examinations, submit answers, and view results automatically through responsive web interfaces. The system is developed using Java, Spring Boot, Hibernate ORM, Thymeleaf, HTML, CSS, JavaScript, and H2/MySQL database technologies. The application follows the MVC (Model-View-Controller) architecture and integrates Spring Security for secure authentication and authorization. Major modules of the system include user registration, login management, exam scheduling, question management, timer-based examinations, automatic answer evaluation, and result generation. Hibernate ORM and JPA are used for efficient database interaction, while Thymeleaf provides dynamic rendering of frontend pages. The main objective of the proposed system is to reduce manual examination work, improve result accuracy, enhance data security, and provide an efficient online examination environment for educational institutions. The application minimizes paperwork, saves administrative effort, improves accessibility, and supports modern digital education systems effectively.

Keywords : Online Examination System, Spring Boot, Hibernate ORM, Java, Thymeleaf, MVC Architecture, Web Application, Automated Evaluation

Author : Sitanshu Malik, Sethi Rohan Kailash, Asst. Prof. Biswajit Sahoo

Title : Online Exam Management System

Volume/Issue : 2026;03(06)

Page No : 250-253

Electronic voting systems have gained significant importance in modern democratic processes due to their speed, accessibility, and efficiency compared to traditional paper-based voting methods. However, existing electronic voting systems face major challenges related to security, privacy, transparency, vote manipulation, and centralized control. Many current e-voting platforms rely on centralized databases, which are vulnerable to cyberattacks, unauthorized modifications, insider threats, and single points of failure. These issues reduce public trust in digital voting systems and raise concerns about election integrity and voter confidentiality.Blockchain technology has emerged as a promising solution for building secure and transparent electronic voting systems. Blockchain provides decentralization, immutability, transparency, and cryptographic security, making it suitable for managing sensitive electoral processes. This project proposes a blockchain-based electronic voting system that ensures voter privacy, vote integrity, and transparent election management using a public blockchain network. The proposed system utilizes cryptographic algorithms, smart contracts, and distributed ledger technology to securely record and verify votes without revealing voter identities.

Keywords : Blockchain, Electronic Voting, Public Blockchain, Smart Contracts, Privacy, Transparency, Cryptography, Cybersecurity, Decentralized Systems.

Author : Kandula Udaybhaskar, G.Vijayasri

Title : SECURE ONLINE VOTING SYSTEM USING BLOCKCHAIN TECHNOLOGY

Volume/Issue : 2026;03(06)

Page No : 254-263

Based Secure Data and Audio Transmission System using UART Protocol is an innovative approach to achieve secure and high-speed data and audio transmission. The system employs visible light signals for communication, providing increased security and reduced electromagnetic interference compared to traditional radio frequency (RF) based methods like Wi-Fi. The system architecture integrates a Li-Fi module with an Arduino microcontroller using the UART protocol, enabling asynchronous serial communication for reliable data transfer. A high-flickering LED array serves as the optical transmitter, while a solar panel or LDR acts as the receiver. The system demonstrates successful simultaneous transmission of hexadecimal text characters and audio signals, displayed on a 16×2 LCD and reproduced through a speaker respectively. The proposed system is suitable for applications requiring high confidentiality, such as military communications, healthcare, and industrial automation.

Keywords : Li-Fi, Visible Light Communication (VLC), UART Protocol, Arduino, LED, Data Transmission, Audio Transmission, Optical Wireless Communication, Photodetector, Embedded Systems.

Author : Dr. P. N. Matte#1, Vishesh Daware#2, Dev Dey#3, Jitesh Chaudhari#4, Pranay Upare#5, Kunal Nande#6

Title : Light-Fidelity Based Text and Audio Data Communication System

Volume/Issue : 2026;03(06)

Page No : 264-270

The Smart IoT Based Medicine Dispenser Robot for Hospitals is an innovative automated system designed to address critical challenges of manual drug dispensing in healthcare environments. This project integrates the ESP-32 microcontroller with a custom-developed mobile application (APK), a Rotary Position Sensor (RPS), Compact Disc (CD) drives, an LCD display, and a mobile robot to create a fully automated and remotely operable medicine dispensing and delivery system. Hospitals face significant risks from medication errors, delayed administration, and overburdened nursing staff. The proposed system mitigates these risks by automating the identification, dispensing, and delivery of prescribed medicines directly to patient bedsides. The mobile APK enables authorized medical personnel to select and schedule medication doses remotely. The ESP-32 processes commands and activates the RPS to identify correct medicine compartments; CD drives release precise doses while the LCD provides real-time feedback. The delivery robot autonomously transports medication to designated wards. The system significantly reduces human error, ensures timely administration, minimizes crosscontamination, and optimizes nursing workloads. With its scalable and costeffective design, this IoT-based solution is particularly suitable for government hospitals, rural healthcare centers, and highpatient-volume wards, offering a robust step toward smart healthcare modernization.

Keywords : IoT, ESP-32, Medicine Dispensing, Hospital Robot, Mobile Application, Rotary Position Sensor, CD Drive Actuator, Healthcare Automation, Smart Hospital, Medication Error Reduction.

Author : Prof. Parag Dhankar#1, Aashlesha Borsare#2 , Vaishnavi Satpute#3 , Gayatri Potdukhe#4 , Kritika Joge#5 , Tushar Mesharam#6 , Samay Hiradave#7

Title : IoT-Based Real-Time MediBot for Medicine Delivery in Hospital Wards

Volume/Issue : 2026;03(06)

Page No : 271-279

Garbage bins remain uncollected for extended periods, posing serious health risks due to potential disease outbreaks, particularly during rainy seasons. Municipal bodies lack automated systems for monitoring garbage levels and notifying central offices in real time. This paper proposes the design and implementation of a GPS and IoT-Based Smart Waste Level and Location Monitoring System that provides real-time information on the fill status and precise geographical location of waste bins. The system integrates an ESP32 microcontroller, ultrasonic sensor for level detection, GPS module (NEO-6M) for location tracking, LCD display for local status indication, buzzer for alerts, and an IoT communication platform for cloudbased monitoring. When the waste level exceeds a configurable threshold, the system transmits the bin's level and GPS coordinates to the IoT server, generating an automated alert for the waste collection personnel. This approach eliminates the need for manual inspection rounds, reduces operational costs, prevents overflow conditions, and promotes timely garbage collection. The system enables centralized, real-time monitoring of multiple distributed waste bins and significantly improves municipal waste management efficiency. Keywords: IoT, Smart Waste Management, ESP32, GPS, Ultrasonic Sensor, Real-Time Monitoring, Municipal Management, Smart Cities, Cloud Server, Overflow Alert.

Keywords :

Author : Dr. Anjum Qureshi#1, Maziya Sheikh#2, Tejal Tupkar#3, Vidhisha Kamble#3, Mamtashri Palakurti#4, Ritika Nagrale#5

Title : IoT-Based Smart Waste Level and Location Monitoring for Efficient Municipal Management

Volume/Issue : 2026;03(06)

Page No : 280-288

Identifying faults in printed circuit board (PCB) electronic circuits is crucial for ensuring the reliability and functionality of electronic devices. Traditional fault identification methods often rely on manual inspection or rule-based techniques, which can be time-consuming and prone to errors. The application of machine learning-based fault identification in PCB electronic circuits has broad implications for various industries, including electronics manufacturing, automotive, aerospace, and consumer electronics. Accurate and timely fault detection enables manufacturers to identify and rectify defects early in the production process, reducing scrap, rework, and warranty costs. Moreover, machine learning models can support predictive maintenance initiatives, allowing for proactive identification of potential failures in electronic systems before they occur. Additionally, fault identification in PCBs facilitates troubleshooting and repair activities, enhancing device reliability and customer satisfaction.Existing methods for fault identification in PCB electronic circuits often rely on manual inspection or limited diagnostic tools, which may be subjective and labor-intensive. These methods may struggle to detect subtle faults or diagnose complex circuit failures, leading to delays in troubleshooting and repair processes. Moreover, traditional approaches may overlook transient or intermittent faults that occur sporadically, resulting in intermittent device malfunctions or failures. Additionally, the increasing complexity and miniaturization of electronic circuits pose challenges for traditional fault identification methods, necessitating more advanced and automated techniques. The proposed system utilizes machine learning techniques to automate and enhance fault identification in PCB electronic circuits, addressing the limitations of existing methods.

Keywords : Printed Circuit Board (PCB), Fault Identification, Machine Learning, Fault Detection, Electronic Circuits, Predictive Maintenance, Automated Inspection, Circuit Diagnostics, Electronic Manufacturing, Reliability Analysis.

Author : Dr.R.K.Krishna#1 , Yash Laxmikant#2 , Yash Shrinivas#3 , Tushar Krushnaji Pandharkar#4 , Karmankar Prince Yuvraj#5

Title : CIRCUIT GUARDIANS: MACHINE LEARNING SOLUTIONS FOR FAULT IDENTIFICATION IN PRINTED CIRCUIT BOARDS

Volume/Issue : 2026;03(06)

Page No : 289-297

Women's safety remains a critical concern in modern society, with increasing incidents of harassment, assault, and health emergencies demanding robust, real-time protective solutions. This paper presents the design and implementation of a Smart Safety Device for Women using IoT, centred on the ESP32 microcontroller. The proposed system continuously monitors vital health parameters including blood oxygen saturation (SpO₂), body temperature, and fall detection, while also integrating a manual panic switch for immediate emergency alerting. Upon detection of any anomaly or panic activation, the system triggers a buzzer alarm, sends an SMS alert with GPS location via a GSM module, and uploads sensor data and a captured image to the IoT cloud platform for remote monitoring. The ESP32-CAM module captures and transmits visual evidence to the cloud server. A 16×2 LCD display provides real-time local feedback. Experimental results demonstrate an overall system response accuracy of 96.8%, with precision of 96.1%, recall of 97.4%, and F1-score of 96.7%. The proposed system outperforms existing wearable and application-based safety devices in terms of real-time monitoring, multi-sensor integration, and autonomous alert capability.

Keywords : Women Safety, IoT, ESP32, GPS, GSM, SpO₂, Fall Detection, Panic Switch, Real-Time Monitoring, Emergency Alert.

Author : Dr. H. S. Pangantiwar#1, Khapne Mayuri Dilip#2 , Salve Madhavi Sanjay#3 , Puddatwar Shreya#4 , Khokle Vaishnavi Ajay#5 , Sayyed Sayma#6 , Chaudhari Vaishnavi#7

Title : SMART SAFETY DEVICE FOR WOMEN USING IOT

Volume/Issue : 2026;03(06)

Page No : 298-306

Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder that affects over 55 million people worldwide, with nearly 10 million new cases reported annually according to the World Health Organization. The disease gradually impairs memory and cognitive abilities, making early detection crucial for effective intervention. Traditionally, Alzheimer’s diagnosis relies heavily on manual assessment of neuroimaging data by neurologists and radiologists, often through visual interpretation of MRI or PET scans. These manual systems are time-consuming, prone to human error, and highly dependent on expert availability. Additionally, manual diagnosis often fails to accurately detect early or intermediate stages, limiting opportunities for timely treatment. Motivated by the need for accurate, rapid, and accessible diagnostic methods, this research proposes an automated system for Alzheimer’s stage classification using MRI images and deep learning techniques, specifically MobileNetV2 integrated with a Random Forest Classifier (RFC). MobileNetV2 efficiently extracts deep and lightweight features from MRI scans, while the RFC enhances classification robustness and generalization. The proposed system preprocesses MRI images, performs deep feature extraction using MobileNetV2, and classifies them into normal, mild cognitive impairment (MCI), and Alzheimer’s stages using RFC. Unlike traditional methods, the approach minimizes subjective bias, reduces diagnostic latency, and improves scalability, making it especially beneficial for rural or underserved healthcare settings. This study aims to enhance diagnostic accuracy while contributing to the development of non-invasive, cost-effective, and intelligent clinical decision support systems. The project integrates key modules including data acquisition, preprocessing, feature extraction, classification, and evaluation, and holds strong potential for real-world deployment in hospitals and research centers for early detection and improved management of Alzheimer’s Disease

Keywords : Alzheimer’s Disease, Structural MRI, Deep Learning, Convolutional Neural Network (CNN), MobileNetV2, Random Forest Classifier (RFC), Early Detection, Mild Cognitive Impairment (MCI), Medical Image Classification, Neurodegenerative Disorders, Feature Extraction, ComputerAided Diagnosis.

Author : Dr.A.P.Sawlikar#1 , Bhoyar Himanshi Anil#2 , Bhagat Tejashri Sanjeev#3 , Mohurle Achal Anil#4 , Pimpalkar Namrata#5 , Lambe Shweta Pravin#6

Title : Convolutional Neural Network for Early Detection of Alzheimer’s Disease Using Structural MRI

Volume/Issue : 2026;03(06)

Page No : 307-315

The behavior of a building during earthquakes depends critically on its overall shape, size and geometry. Earthquake resistant design of buildings depends upon providing the building with strength, stiffness and inelastic deformation capacity which are great enough to withstand a given level of earthquake generated force. This is generally accomplished through the selection of an appropriate building configuration and the careful detailing of structural members. Configuration is critical to good seismic performance of buildings. The important aspects affecting seismic configuration of buildings are overall geometry, structural systems, and load paths. The building slenderness ratio and the building core size are the key drivers for the efficient structural design. This paper focuses on the effect of Slenderness Ratio of the building frame, B is the Base width and L is the Length of the building frame with different Plan Configurations on the Seismic Analysis of Multistoried Regular R.C.C. Buildings. The slenderness ratios assumed are 3.00, 3.60, 4.80 and 7.20 with varying building widths as 25m, 20m, 15m and 10m. The analysis results like storey drift, storey shear, storey bending, building torsion, time period, model stiffness values are compared with all the aspect ratios for G+23 building by using ETABS Software.

Keywords : Earthquake, slenderness ratio, storey drift, storey shear, storey bending, building torsion, time period, model stiffness.

Author : SUNKARA PADMA TEJA1 E SRINIVAS2

Title : COMPARATIVE ANALYSIS OF G+23 BUILDING WITH VARIOUS SLENDERNESS RATIOS BY USING ETABS SOFTWARE

Volume/Issue : 2026;03(06)

Page No : 316-322

Heart disease remains one of the leading causes of mortality worldwide, demanding accurate and early diagnosis systems. Traditional diagnostic approaches rely on clinical expertise and structured medical data; however, realworld healthcare data is frequently contaminated by noise from sensor errors, missing values, and environmental disturbances. This paper proposes a robust heart sound classification system leveraging Convolutional Neural Networks (CNN) integrated with advanced preprocessing techniques to handle noisy input data effectively. The system classifies heart sounds as either Healthy or Abnormal using audio features such as Mel-Frequency Cepstral Coefficients (MFCC) extracted from raw phonocardiogram recordings. Experimental evaluation on a benchmark dataset demonstrates that the CNN-based model achieves 96.8% accuracy, significantly outperforming conventional Machine Learning (ML) algorithms including K-Nearest Neighbor (KNN), Decision Tree, Random Forest, Support Vector Machine (SVM), and Artificial Neural Network (ANN). The proposed approach provides a scalable, noise-resilient diagnostic framework suitable for real-time healthcare and IoT-based medical monitoring applications.

Keywords : Heart Sound Classification, CNN, Deep Learning, Machine Learning, MFCC, Noise Reduction, Healthcare AI, ECG, Phonocardiogram, Signal Processing

Author : Kampati Nagadevi1 , N. V. Ashok Kumar2 , C. P. V. N. J. Mohan Rao3

Title : HEALTHY AND ABNORMAL HEART SOUND CLASSIFICATION USING MACHINE LEARNING AND DEEP LEARNING TECHNIQUES

Volume/Issue : 2026;03(06)

Page No : 323-329

Financial technology (FinTech) has become an urgent topic with regard to financial fraud, because of the numerous online transactions, and the changing patterns of online fraud. To improve the effectiveness and robustness of fraud detection, this research suggests an adaptive machine learning (ML)-based fraud detector. The methodology entails using the IEEE-CIS fraud detection dataset, which is preprocessed, feature-engineered, and treated with classimbalance remedies like SMOTE. LightGBM and XGBoost are two modernized ensemble models, that are implemented and tested. Explainable AI models, such as SHAP and LIME, are used to increase the model transparency. Experimental results indicate that XGBoost is more effective than LightGBM, achieving higher accuracy (99.1%) as well as higher precision, recall and F1-score. The given structure is a good way to capture the intricate patterns of fraud without becoming uninterpretable. This study demonstrates the significance of real-time fraud detection using adaptive, scalable, explainable ML solutions in current financial systems.

Keywords : Financial Technology (FinTech), Fraud Detection, Machine Learning, Explainable AI, Financial Transactions.

Author : Vikas Kumar Pandey

Title : Adaptive Fraud Detection Based on Machine Learning Models for Financial Technology Applications

Volume/Issue : 2026;03(06)

Page No : 330-337

Performance appraisal systems play a crucial role in evaluating employee performance, enhancing productivity, and supporting organizational growth. This study examines the effectiveness of the Employee Performance Review System at Tech Mahindra, a leading global information technology and consulting organization. The research aims to evaluate the existing appraisal process and understand employees' perceptions regarding its fairness, transparency, and accuracy. The study further investigates the methods and procedures adopted in performance evaluations and their impact on employee motivation, career development, and reward management. Data were collected from employees through a structured questionnaire and analyzed using descriptive statistical techniques. The findings indicate that employees generally perceive the appraisal system positively, particularly in terms of feedback, performance recognition, and developmental opportunities. However, certain areas such as communication of evaluation criteria, consistency in assessments, and employee participation require improvement. The study highlights the significance of an effective appraisal system in aligning individual performance with organizational objectives. It also emphasizes the role of continuous feedback, transparent evaluation mechanisms, and employee involvement in enhancing appraisal effectiveness. The results suggest that a well-designed performance review system contributes to higher employee satisfaction, motivation, and organizational commitment. The study concludes that strengthening appraisal practices can support employee growth and improve overall organizational performance at Tech Mahindra.

Keywords : Performance Appraisal System, Employee Performance, Employee Motivation, Career Development, Tech Mahindra

Author : Dr. I. M. Christina Febiula

Title : EXAMINING EMPLOYEE PERFORMANCE REVIEW SYSTEMS: EVIDENCE FROM TECH MAHINDRA

Volume/Issue : 2026;03(06)

Page No : 338-350

Digital banking adoption has accelerated globally, increasing both transaction volumes and the risk of fraudulent activities. Traditional fraud detection techniques, relying on batch processing and periodic analysis, fail to detect fraudulent transactions in real time, leading to financial losses and customer distrust. This paper proposes a real-time bank transaction fraud detection system that leverages Apache Kafka for streaming transaction data and integrates machine learning algorithms such as Random Forest and Naive Bayes for real-time classification. The system architecture ensures scalability, fault tolerance, and low-latency processing. Experimental evaluation on a large-scale bank transaction dataset demonstrates that the proposed system achieves high accuracy, precision, and recall, efficiently detecting fraudulent transactions while minimizing false positives. The framework provides a practical solution for modern banking institutions seeking proactive fraud management.

Keywords :

Author : CH.Sai Sushmitha, K.Subash Chandra

Title : REAL-TIME BANK TRANSACTION FRAUD DETECTION USING KAFKA ALGORITHM

Volume/Issue : 2026;03(06)

Page No : 361-365

Liver diseases represent a major global health challenge due to their asymptomatic nature in early stages and high mortality rates when detected late. Accurate and early diagnosis is crucial for effective treatment and improved patient outcomes. This paper proposes a hybrid intelligent system that integrates Genetic Algorithm (GA) for feature selection and optimization with Convolutional Neural Networks (CNN) for automated classification of liver diseases. The Genetic Algorithm enhances the model performance by selecting the most relevant features from medical datasets, thereby reducing dimensionality and improving computational efficiency. The CNN model is employed to learn complex patterns from imaging or structured clinical data. The proposed system demonstrates improved accuracy, sensitivity, and specificity compared to traditional machine learning approaches. Experimental results validate the robustness and reliability of the hybrid model in diagnosing liver diseases effectively.

Keywords :

Author : CH.Sai Sushmitha, B.Bhagya Lakshmi

Title : DIAGNOSIS OF LIVER DISEASES USING GENETIC ALGORITHM AND CNN

Volume/Issue : 2026;03(06)

Page No : 366-369

Online payment systems have become a cornerstone of digital finance, offering convenience and efficiency. However, these systems are highly susceptible to fraudulent transactions, resulting in significant financial losses annually. Detecting fraud is challenging due to the highly imbalanced nature of transaction data, where fraudulent activities represent a very small portion of all transactions. Traditional rule-based detection systems often fail to capture evolving fraud patterns. This paper proposes an intelligent fraud detection framework leveraging Random Forest and Naïve Bayes classifiers combined with Synthetic Minority Over-sampling Technique (SMOTE) to mitigate data imbalance. The framework involves preprocessing transaction data, balancing the dataset using SMOTE, and applying machine learning classifiers to predict fraudulent transactions. Experimental evaluation demonstrates that Random Forest, in particular, provides superior accuracy, precision, recall, and F1-score. This study highlights the importance of ensemble and probabilistic approaches, combined with oversampling techniques, for effective online fraud detection.

Keywords :

Author : D.Sunitha,I.Haritha

Title : ONLINE FRAUD PAYMENT DETECTION USING RANDOM FOREST AND NAÏVE BAYES WITH SMOTE

Volume/Issue : 2026;03(06)

Page No : 370-374

The exponential growth of online transactions has led to a significant rise in credit card fraud, causing substantial financial losses to financial institutions and customers. Detecting fraudulent transactions is a challenging task due to the high volume of transaction data and the evolving nature of fraud patterns. This paper presents a machine learning-based approach for credit card fraud detection using the Random Forest algorithm. The proposed system analyzes historical transaction data and classifies transactions as genuine or fraudulent. Various preprocessing techniques such as data cleaning, normalization, and feature extraction are applied to enhance model performance. The system is evaluated using performance metrics such as accuracy, precision, recall, and F1-score. Experimental results demonstrate that the Random Forest algorithm provides superior performance compared to traditional methods, achieving high accuracy and robustness in fraud detection.

Keywords :

Author : D.Navya Sai Lakshmi, M.Anuradha

Title : CREDIT CARD FRAUD DETECTION USING MACHINE LEARNING

Volume/Issue : 2026;03(06)

Page No : 375-380

The identification of lung cancer at the early stage is very demanding and difficult task due to construction of the cell. The cancer grows in the body when cancerous cells start to develop uncontrollably. The image processing plays vital role in the prediction of lung cancer at early stage which is also helpful in treatment to avoid the lung cancer. This proposed system is developed to detect lung cancer at early stage with the help of image processing techniques and artificial neural network classifier to design computer based diagnosis system. In this system, during the preprocessing step, several image enhancing techniques, masks are applied using morphological operations and thresholding technique, which eliminates background and surrounding tissue. Region of interest (ROI) is calculated using region based segmentation algorithm. Circle fit algorithm is used to extract the desired nodule. Radius, Mean Intensity, Area, Euler Number and ECD features are extracted in feature extracting step. Finally, Back propagation algorithm is used to train Artificial Neural Network (ANN) in categorization stage.v

Keywords :

Author : G.Neehiraka,S.Jaya Sri

Title : A REVIEW ONEARLY DETECTION OF CANCER USING AI

Volume/Issue : 2026;03(06)

Page No : 381-387

The growing digital landscape has intensified the need for robust data security measures to safeguard sensitive information from cyber threats. This paper explores the integration of blockchain technology and artificial intelligence (AI) as a synergistic approach to enhance data security. Blockchain, known for its decentralized and tamper-resistant nature, provides a secure foundation for storing and managing data. AI, with its advanced analytical capabilities, adds an intelligent layer to identify and respond to potential threats. Blockchain's decentralized architecture eliminates a single point of failure, reducing the risk of unauthorized access or data manipulation. Each block in the chain contains a cryptographic hash of the previous block, creating an immutable and transparent ledger. This ensures data integrity and establishes a reliable audit trail for all transactions, bolstering overall security. AI complements blockchain by enabling real-time threat detection and adaptive response mechanisms. Machine learning algorithms analyze patterns and anomalies within the data, identifying potential security breaches or abnormal activities. Through continuous learning, AI systems evolve to adapt to new and sophisticated threats, enhancing the overall resilience of the security infrastructure. The integration of blockchain and AI introduces a novel paradigm for data security. Smart contracts, selfexecuting contracts with encoded rules on the blockchain, automate security protocols and access controls, minimizing human error and reducing vulnerabilities. Additionally, AI-driven predictive analytics can forecast potential security risks, allowing proactive measures to be implemented. This paper discusses case studies and applications where the combined use of blockchain and AI has proven effective in securing various data-intensive domains, such as finance, healthcare, and supply chain management. The collaborative efforts of these technologies create a robust defense against evolving cyber threats, ensuring data confidentiality, integrity, and availability in an increasingly interconnected digital world. As organizations strive to fortify their data security measures, the integration of blockchain and AI emerges as a powerful solution to address the dynamic challenges of the modern cybersecurity landscape.

Keywords :

Author : G.Poojitha Devi, D.Daiva Kumari

Title : SECURING DATA WITH BLOCKCHAIN AND AI

Volume/Issue : 2026;03(06)

Page No : 388-395

In recent times, there has been a drastic change in people’s lifestyles and with an increase in incomes and lower cost of automobiles there is a huge increment in the number of cars on the roads which has led to traffic and commotion. The manual efforts to keep people from breaking traffic rules such as the speed limit are not enough. There is not enough police and man force available to track the traffic and vehicles on roads and check them for speed control. Hence, we require technologically advanced speed calculators installed that effectively detect cars on the road and calculate their speeds. To implement the above idea two basic requirements, need to be met which are the effective detection of the cars on roads and their velocity measurement. For this purpose, we can use OpenCV software which uses the Haar cascade to train our machine to detect the object, in this case the car. we have developed a Haar cascade to detect cars on the roads, whose velocities are then measured using a python script. The real-time application of this project proves to be much useful as it is easy to implement, fast to process and efficient with low cost development. Also, the tool might be useful to apply in simulation tools to measure velocities of cars. This can be further developed to identify all kinds of vehicles as well as to check anyone who breaks a traffic light. The improvements in the project can be done by creating a bigger haar cascade since bigger the haar cascade developed, more the number of vehicles that can be detected on the roads. Better search algorithms can allow a faster search and better detection of these vehicles for better efficiency

Keywords :

Author : K.Kavya Sri, Dr P.Chiranjeevi

Title : VEHICLE TRACKING AND SPEED ESTIMATION

Volume/Issue : 2026;03(06)

Page No : 396-401

The proliferation of counterfeit currency poses a significant threat to economic stability and security. In response to this challenge, this research proposes a robust solution for Fake Currency Identification utilizing Convolutional Neural Networks (CNNs). CNNs, a class of deep learning models, have demonstrated remarkable success in image recognition tasks, making them an ideal choice for the complex and nuanced patterns found in currency notes. The proposed system begins with a comprehensive dataset comprising genuine and counterfeit currency images, capturing a diverse range of features and variations. The CNN architecture is designed to automatically learn and extract intricate patterns and features crucial for distinguishing authentic and fake banknotes. The model's training involves optimizing its parameters through iterative processes, enhancing its ability to generalize and identify subtle differences in visual characteristics. To achieve effective feature extraction, the CNN utilizes multiple convolutional layers, pooling layers, and fully connected layers. The trained model demonstrates a high degree of accuracy in discriminating between real and counterfeit currencies. Moreover, transfer learning techniques may be employed to leverage pre-trained CNN models on larger datasets, further enhancing the system's performance. The implementation also considers real-world scenarios, including variations in lighting conditions, orientations, and possible image distortions. The robustness of the system is validated through extensive testing, showcasing its ability to adapt and accurately identify fake currency in diverse environments. The proposed Fake Currency Identification system using CNNs contributes to the ongoing efforts to combat financial fraud. Its automated and accurate nature offers an efficient means for financial institutions, businesses, and law enforcement agencies to detect counterfeit currency, thereby safeguarding the integrity of monetary systems. Additionally, the adaptability of the model allows for potential integration into existing security frameworks, providing a scalable and effective solution against the constant evolution of counterfeit practices.

Keywords :

Author : M.Rupa Devi, Dr G.Vijaya Kumar

Title : FAKE CURRENCY IDENTIFICATION

Volume/Issue : 2026;03(06)

Page No : 402-408

Crowd counting plays a crucial role in various applications, such as urban planning, event management, and public safety. Traditional methods for crowd counting often face challenges in accuracy and efficiency, prompting a shift towards machine learning techniques. This abstract provides a comprehensive overview of recent advancements in crowd counting using machine learning. Machine learning models, particularly convolutional neural networks (CNNs) and their variants, have shown remarkable success in handling the complexities of crowd counting. These models leverage their ability to automatically learn intricate patterns and features from images, enabling more accurate and robust crowd estimation. The utilization of deep learning architectures facilitates the extraction of hierarchical features, allowing for better representation of crowded scenes. This review discusses the diverse approaches employed in crowd counting, encompassing both supervised and unsupervised learning paradigms. Supervised methods rely on annotated datasets for model training, while unsupervised methods explore novel ways to estimate crowd density without labeled data. Additionally, semi-supervised techniques leverage a combination of labeled and unlabeled data to enhance model performance. The challenges associated with crowd counting, such as scale variations, occlusions, and diverse crowd behaviors, are addressed in the context of machine learning. Various data augmentation strategies, regularization techniques, and attention mechanisms are explored to improve model generalization and robustness. Transfer learning is also discussed as a means to adapt pre-trained models to new crowd counting scenarios, reducing the need for large annotated datasets. Furthermore, the review highlights the integration of crowd counting models with real-world applications, including smart cities, transportation systems, and public safety. The ethical considerations related to crowd surveillance and privacy are discussed, emphasizing the importance of responsible deployment of machine learning in crowd counting applications. In conclusion, this abstract provides a comprehensive overview of the state-of-the-art in crowd counting using machine learning, emphasizing the advancements, challenges, and practical applications. The integration of machine learning techniques into crowd counting holds great promise for addressing the complexities of real-world scenarios and advancing the field towards more accurate and efficient crowd estimation systems.

Keywords :

Author : M.Meghana, B.Bhagya Lakshmi

Title : CROWD COUNTING USING MACHINE LEARNING

Volume/Issue : 2026;03(06)

Page No : 409-417

Traffic congestion is becoming one of the critical issues with increasing population and automobiles in cities. Traffic jams not only cause extra delay and stress for the drivers, but also increase fuel consumption and air pollution. Although it seems to pervade everywhere, megacities are the ones most affected by it. And its ever-increasing nature makes it necessary to calculate the road traffic density in real-time for better signal control and effective traffic management. The traffic controller is one of the critical factors affecting traffic flow. Therefore, the need for optimizing traffic control to better accommodate this increasing demand arises. Our proposed system aims to utilize live images from the cameras at traffic junctions for traffic density calculation using image processing and AI. It also focuses on the algorithm for switching the traffic lights based on the vehicle density to reduce congestion, thereby providing faster transit to people and reducing pollution.

Keywords :

Author : P.Deepa sree, M.Anuradha

Title : SMART CONTROL OF TRAFFIC LIGHT USING ARTIFICIAL INTELLIGENCE

Volume/Issue : 2026;03(06)

Page No : 418-425