Journal of Science Engineering Technology and Management Science

Twitter has become a prominent platform for real-time information exchange, with over 500 million tweets generated daily, a substantial proportion of which are related to dynamic real-world events. Political incidents alone account for nearly 40% of trending discussions worldwide, making automated event monitoring essential for timely decision-making and crisis response. Manual classification of such high-volume textual data is labor-intensive, inconsistent, and impractical for large-scale or real-time analysis. Furthermore, conventional machine learning techniques often fail to capture contextual and semantic nuances necessary to distinguish diverse event categories accurately. To address these limitations, this study proposes an automated political event monitoring framework that integrates transformer-based embeddings with optimized classification strategies. The methodology begins with comprehensive Natural Language Processing (NLP) preprocessing, including tokenization, stop-word removal, normalization, and lemmatization, followed by Exploratory Data Analysis (EDA) to identify trends and data distributions. Context-aware semantic features are extracted using Lightweight Robustly Optimized Bidirectional Encoder Representations from Transformers (RoBERTa), enabling efficient yet rich textual representations. Unlike baseline classifiers such as standard Stochastic Gradient Descent (SGD), Histogram-based Gradient Boosting (HGB), Greedy Tree Classifier (GTC), and Random Forest Classifier (RFC), the proposed approach incorporates Deep Neural Network (DNN)-based feature selection combined with an optimized SGD classifier to enhance discriminative learning and handle class imbalance effectively. The system categorizes tweets into six event classes: disaster, political, positive, protest, riot, and terror. Experimental results demonstrate improved accuracy, scalability, and adaptability, making the framework suitable for real-time political event detection and supporting applications in governance, public safety, and crisis management.

Keywords : Twitter data, Data mining, Natural Language Processing, Transformer-based embeddings, BERT architecture, Deep learning.

Author : S. Swapna, Keesari Rajashekhar, Jallika Rakesh, Mohammad Khaleel Pasha, Kogila Lavan Sai, Mohammad Siraj

Title : A RoBERTa-Driven Deep Feature Learning Framework for Real-Time Political Event Classification on Twitter

Volume/Issue : 2026;03(03)

Page No : 245-257

Healthcare systems are increasingly vulnerable to cyberattacks, with over 93 million patient records breached in 2023 and an annual increase of 22% in vulnerability disclosures affecting hospital networks. Manual approaches to bug severity detection are error-prone and inefficient, often failing to manage the growing volume of unstructured medical security reports. To address this, we propose a Natural Language Processing (NLP) framework that leverages a Medical NLP dataset comprising incident reports, advisories, and vulnerability disclosures. The methodology begins with NLP preprocessing and Exploratory Data Analysis (EDA) to clean, normalize, and visualize the dataset. Following this, Lightweight RoBERT (Robustly Optimized BERT) is used with word embeddings for semantic representation of text, ensuring reduced computational cost compared to heavy transformer models. Unlike existing systems based on Stochastic Gradient Descent (SGD) and Natural Gradient Boosting (NGBoost) Classifier, the proposed pipeline integrates Deep Neural Network (DNN) based feature selection with NGBoost to improve predictive performance. The system classifies bug severity into two categories: Normal and Severity, providing actionable insights for prioritizing critical vulnerabilities. By combining contextual embeddings with advanced feature selection, the proposed approach enhances accuracy, reduces misclassification, and ensures faster real-time response. This innovation contributes to strengthening the cybersecurity posture of the healthcare ecosystem by delivering an efficient and scalable solution for automated bug severity detection.

Keywords : Healthcare system, Cyber-attack detection, Natural Language Processing, BERT architecture, Deep neural networks, Natural gradient boosting classifier.

Author : Ayesha Banu, Valluru Bhavani, Pooduru Akash, Raparthi Yadeesh, Thumma Swathi

Title : Automated Bug Severity Prediction in Healthcare Security Text using Lightweight RoBERTa and Ensemble Learning

Volume/Issue : 2026;03(03)

Page No : 258-271

Technological advancements have fundamentally transformed modern defense operations, particularly within surveillance, reconnaissance, and battlefield decision support systems. While military environments generate vast volumes of visual data from satellites and unmanned aerial vehicles (UAVs), conventional manual interpretation remains resource-intensive and prone to human error, hindering real-time operational responses. To address these systemic bottlenecks, this research introduces an intelligent, high-assurance decision support framework for the automated classification of tactical military imagery. The proposed system transitions beyond traditional rule-based software by integrating a suite of soft computing models, including Perceptron, Decision Tree Classifiers (DTC), and Deep Neural Networks (DNN). Central to the framework is a novel Hybrid Convolutional Recurrent Model (CRM), which synergizes Convolutional Neural Networks (CNN) for spatial feature extraction with Long Short-Term Memory (LSTM) networks to capture essential temporal dependencies in dynamic battlefield scenarios. The architecture is encapsulated within a modular graphical interface designed for streamlined data ingestion, model training, and performance visualization. Experimental validation demonstrates that the integrated CRM significantly enhances processing speed and classification reliability, providing a scalable and robust technological solution for modern military intelligence and tactical decision-making.

Keywords : Tactical decision support, Convolutional recurrent model, Military environments, surveillance sensors

Author : J. Sravanthi, Polsani Spoorthi, Tharala Raghavi, Shaik Asif, Burra Sai Vinay

Title : Intelligent Military Decision Support Using Machine Learning and DataDriven Tactical Analysis

Volume/Issue : 2026;03(03)

Page No : 272-281

Project documentation plays a critical role in effective project management, yet extracting relevant information from large and complex PDF documents remains a challenging and timeconsuming task. Project managers and team members often spend significant effort manually searching through documents such as requirement specifications, project plans, risk registers, and status reports to locate essential information. Traditional project management tools mainly focus on task tracking and collaboration but lack advanced capabilities for intelligent document understanding and contextual retrieval. To address these limitations, this research proposes an AIpowered Project Management Assistant Chatbot that leverages Retrieval-Augmented Generation (RAG) architecture to enable efficient information extraction and conversational interaction with project documents. The system processes uploaded PDF files by converting them into machinereadable text, performing semantic chunking, and generating vector embeddings using advanced natural language processing models. These embeddings are stored in a vector database that enables fast and contextually relevant document retrieval. When users ask questions, the system retrieves the most relevant document sections and provides accurate responses through a large language model. The chatbot interface allows project managers and team members to interact with project documentation through natural language queries, significantly reducing the time required to locate information. The proposed system enhances decision-making, improves knowledge accessibility, and supports efficient project monitoring. Experimental observations demonstrate that the system can effectively retrieve context-relevant information and provide meaningful answers from project documents. This approach contributes to intelligent document management in project environments by combining natural language processing, semantic search, and conversational AI technologies.

Keywords :

Author : Dr.G. SYAM PRASAD,R.L.V.N.S. LAVANYA, V. GEETHA LATHA, G. RAJESWARI, M. BHASKARA RAO

Title : PROJECT MANAGEMENT ASSISTANT AI CHATBOT USING PDF

Volume/Issue : 2026;03(03)

Page No : 282-289

Heart disease remains one of the leading causes of mortality worldwide, creating an urgent need for accurate and early diagnostic systems that can support healthcare professionals in clinical decision-making. Traditional diagnostic procedures often rely on manual interpretation of clinical parameters such as electrocardiogram readings, cholesterol levels, blood pressure, and patient medical history. While these approaches are effective in many cases, they may fail to capture complex nonlinear relationships among multiple risk factors and may also be affected by time constraints and subjective interpretation. Recent advancements in machine learning provide promising opportunities for developing automated and intelligent diagnostic systems capable of analyzing large volumes of healthcare data with improved accuracy and efficiency. This research proposes a heart disease classification system using the Extreme Gradient Boosting (XGBoost) algorithm to predict the presence of cardiovascular disease based on clinical attributes. The proposed model incorporates several stages including data preprocessing, feature engineering, model training, and performance evaluation. The system utilizes structured medical datasets containing various patient attributes such as age, gender, chest pain type, cholesterol level, blood pressure, fasting blood sugar, and electrocardiographic results. XGBoost is selected due to its ability to handle missing data, manage nonlinear relationships, and optimize predictive performance through gradient boosting techniques. Experimental evaluation demonstrates that the proposed system achieves high predictive accuracy while maintaining strong precision, recall, and F1-score values. The model can assist clinicians by providing early risk prediction and supporting informed medical decisions. Overall, the proposed approach contributes to the advancement of AI-based healthcare systems by enabling efficient, scalable, and data-driven heart disease prediction mechanisms.

Keywords :

Author : Dr.P. SAMBASIVA RAO, A. ARUN KUMAR, K. DENI SRI SAI LALITHA, M. ABHINAYA SRI, SK.VENU BABU

Title : HEART DISEASE CLASSIFICATION USING XGBOOST CLASSIFIER

Volume/Issue : 2026;03(03)

Page No : 290-298

Road infrastructure plays a critical role in transportation efficiency, economic development, and public safety. However, road defects such as potholes, cracks, and surface deformations significantly impact vehicle safety and increase maintenance costs. Traditional road inspection methods rely on manual surveys conducted by engineers or maintenance personnel, which are time-consuming, labor intensive, and prone to human error. Automated defect detection using computer vision and machine learning has emerged as a promising solution to address these challenges. This study proposes an intelligent road defect detection system that integrates Long Short-Term Memory (LSTM) networks with a Random Forest classifier to improve the accuracy and reliability of road surface condition monitoring. The system processes road surface images and sequential data captured through cameras or sensors to identify different types of road defects. LSTM networks are used to analyze sequential patterns and temporal dependencies in road surface data, enabling the detection of progressive deterioration over time. Random Forest algorithms are utilized for efficient classification of extracted features and defect categories such as cracks, potholes, and surface irregularities. The hybrid approach combines temporal learning capabilities with robust classification techniques, improving detection accuracy and reducing false predictions. Feature extraction techniques are applied to capture texture, shape, and intensity variations in road images before feeding them into the classification models. Experimental results demonstrate that the proposed hybrid model achieves higher accuracy and reliability compared with traditional imageprocessing methods and single-model approaches. The system supports early identification of road defects and provides valuable insights for proactive road maintenance. Ultimately, the proposed framework contributes to the development of smart transportation infrastructure by enabling automated, scalable, and cost-effective road condition monitoring systems.

Keywords :

Author : Dr.M. NANCHARAIAH, K. HARIKA, G. RADHA KRISHNA, B. ANUSHA, B. DOLA DEEPAK

Title : ROAD DEFECT DETECTION USING ISTM AND RANDOM FOREST

Volume/Issue : 2026;03(03)

Page No : 299-307

Ovarian cancer is one of the most lethal gynecological malignancies worldwide due to its silent progression and late-stage diagnosis. Early detection significantly improves survival rates; however, conventional diagnostic approaches often fail to identify the disease at an initial stage. This study presents an Artificial Intelligence (AI)-driven Clinical Decision Support System (CDSS) designed to predict ovarian cancer risk using clinical and biochemical data. The proposed system integrates machine learning and deep learning algorithms to analyze multiple patient parameters, including age, menopausal status, hematological indicators, metabolic biomarkers, and tumor markers such as CA125, HE4, and AFP. The dataset is preprocessed through normalization, missing value handling, and feature selection to improve model accuracy. Various predictive models such as Logistic Regression, Support Vector Machines (SVM), Random Forest, and Deep Neural Networks are implemented and compared to identify the most reliable predictive model. Additionally, Explainable Artificial Intelligence (XAI) techniques, particularly SHAP (SHapley Additive Explanations), are employed to enhance transparency and interpretability by identifying the contribution of each biomarker in predicting cancer risk. The system architecture incorporates a scalable web-based platform with a FastAPI backend, a React frontend, and integrated database support for clinical data storage and user management. This enables healthcare professionals to input patient parameters and receive real-time predictions along with interpretable insights. Experimental results demonstrate that the proposed model achieves improved predictive performance compared to traditional diagnostic approaches. The developed system has the potential to assist clinicians in early screening, risk assessment, and decision-making, thereby improving early diagnosis and patient outcomes in ovarian cancer management.

Keywords :

Author : Prof.S.V.C. GUPTA,L. MOHANA JHANSI,K. LALITHA SUDHA RANI,J.UMA DEVI,MD.IBRAHIM

Title : OVARIAN CANCER PREDICTION

Volume/Issue : 2026;03(03)

Page No : 308-316

The rapid growth of social media platforms has significantly transformed the way information is produced, shared, and consumed across the world. While these platforms provide a powerful medium for communication and information exchange, they also facilitate the rapid spread of fake news and disinformation. Such misinformation can influence public opinion, disrupt political processes, and create social instability. Traditional fact-checking mechanisms rely heavily on manual verification, which is time-consuming, resource-intensive, and incapable of handling the massive volume of online content generated daily. Consequently, automated approaches using Machine Learning (ML) and Natural Language Processing (NLP) have emerged as promising solutions for identifying and mitigating fake news. This study presents a systematic review and comparative analysis of machine learning and deep learning techniques used for fake news detection. In particular, the research focuses on the performance comparison between Random Forest, a widely used machine learning algorithm, and Long Short-Term Memory (LSTM), a deep learning architecture designed for sequential text processing. Random Forest utilizes textual features extracted through TF-IDF vectorization, while LSTM leverages word embeddings to capture contextual semantics and long-term dependencies within textual data. The proposed framework includes data collection from benchmark datasets, text preprocessing, feature extraction, model training, and evaluation using performance metrics such as accuracy, precision, recall, and F1-score. The results demonstrate that deep learning models capture contextual patterns more effectively, whereas traditional machine learning models provide faster training and interpretability. The study highlights the importance of combining linguistic features with deep contextual learning to improve misinformation detection. Furthermore, it identifies research gaps and future opportunities, including multilingual detection systems, transformer-based models, and real-time deployment for social media monitoring

Keywords :

Author : N. ANIL KUMAR, M. BHAVYA SREE, M. LAKSHMI GANESH, CH.SYAM VITAL KUMAR, FARHAT SULTANA

Title : SYSTEMATIC REVIEW ON FAKE NEWS & DISINFORMATION USING ML

Volume/Issue : 2026;03(03)

Page No : 317-324

Ad click fraud has become a major challenge in the digital advertising ecosystem, leading to significant financial losses for advertisers and reducing the reliability of online advertising metrics. Fraudulent clicks are often generated using automated bots, click farms, or malicious scripts designed to inflate advertising revenue or exhaust competitors’ advertising budgets. With the rapid growth of payper-click advertising models, detecting and preventing such fraudulent activities has become critical. Traditional rule-based detection methods often fail to identify sophisticated fraud patterns due to their inability to adapt to evolving attack strategies. Machine Learning (ML) and Deep Learning (DL) techniques provide more effective solutions by analyzing large volumes of user interaction data and identifying abnormal behavioral patterns. This study proposes a fraud detection framework that utilizes supervised machine learning algorithms and deep neural networks to detect suspicious click patterns in real time. The system processes historical clickstream data, user interaction logs, and network attributes to train predictive models capable of distinguishing legitimate clicks from fraudulent ones. Feature engineering techniques are applied to capture temporal patterns, device information, IP behavior, and session characteristics. The proposed framework combines classification algorithms such as Random Forest, Support Vector Machine, and Gradient Boosting with deep learning architectures like Artificial Neural Networks to improve detection accuracy. Experimental evaluation demonstrates that the hybrid approach significantly enhances fraud detection performance compared with traditional approaches. The system can assist digital advertising platforms in minimizing revenue loss, improving transparency, and ensuring fair advertising practices. The results highlight the importance of intelligent automated fraud detection systems in safeguarding the integrity of the online advertising ecosystem

Keywords :

Author : V.M.R. KRISHNA RAO, K. KEERTHI KALA, P. KAVITHA, CH.BALADITYA, D. NAVEEN

Title : AD CLICK FRAUD DETECTION WITH ML AND DL

Volume/Issue : 2026;03(03)

Page No : 325-333

Energy consumption forecasting has become an essential component of modern power system management due to increasing global energy demand, rapid urbanization, and the growing complexity of electricity distribution networks. Accurate forecasting enables utilities and energy providers to optimize resource allocation, reduce operational costs, improve grid stability, and support sustainable energy planning. Traditional statistical techniques such as autoregressive integrated moving average models and exponential smoothing methods have been widely used for load prediction; however, these methods often struggle to capture nonlinear patterns and dynamic relationships present in real-world energy consumption data. In recent years, machine learning techniques have demonstrated superior capabilities in modeling complex temporal patterns and extracting hidden relationships from large datasets. This study presents a machine learning– based energy consumption forecasting system using the Random Forest Regressor algorithm. The proposed system utilizes historical energy usage data along with relevant environmental and temporal features such as temperature, humidity, time of day, day of week, and seasonal indicators to improve prediction accuracy. The system architecture includes data preprocessing, feature engineering, model training, evaluation, and prediction modules. Random Forest was selected due to its ensemble learning capability, robustness against overfitting, and ability to handle nonlinear relationships and noisy data. Experimental evaluation demonstrates that the proposed model produces reliable forecasting performance and can assist energy providers in demand planning and operational decision-making. The implementation of such intelligent forecasting systems can significantly enhance energy efficiency, support renewable energy integration, and contribute to the development of smart grid infrastructure.

Keywords :

Author : P. ASHOK KUMAR, A. BHAVYA SRI, K. ALEKHYA, K. RUCHITHA, K. KEERTHANA SAI

Title : ENERGY CONSUMPTION FORECASTING USING ML

Volume/Issue : 2026;03(03)

Page No : 334-342

The rapid advancement of Artificial Intelligence (AI) and Machine Learning (ML) has significantly transformed the healthcare sector by enabling predictive analytics, early disease detection, and data-driven clinical decision-making. However, healthcare institutions often face challenges in sharing sensitive patient data due to privacy regulations, security risks, and ethical concerns. Traditional centralized machine learning approaches require aggregating data from multiple organizations, which increases the risk of data breaches and violates strict privacy policies. To address these issues, this study proposes a privacypreserving federated learning platform integrated with Large Language Model (LLM) automation for intelligent healthcare risk prediction. The proposed system allows multiple healthcare institutions to collaboratively train a global machine learning model without sharing raw patient data. Instead, each institution performs local model training and transmits only model parameters to a central aggregation server. These parameters are combined using the Federated Averaging algorithm to produce a generalized global predictive model while maintaining data privacy. Furthermore, the system incorporates a transformer-based LLM to generate human-readable explanations for prediction results, improving transparency and interpretability for medical professionals. The platform is implemented using a full-stack architecture consisting of a React-based frontend, Node.js backend, MongoDB database, and a Python Flask microservice for machine learning operations. The system also includes modules for authentication, dataset management, federated training coordination, model performance monitoring, and AI-driven explanation generation. Experimental results demonstrate that the proposed approach improves predictive accuracy while maintaining strict data privacy standards. The integration of federated learning and explainable AI provides a scalable and secure framework for collaborative healthcare analytics.

Keywords :

Author : K. VENKATESWARA RAO, S. PRIYA VASANTHI, J. CHANDINI,R. CHINNI HEMADRI KUMAR, MD.ASFIN

Title : FEDERATED LEARNING WITH LLM AUTOMATION - WEB-BASED SYSTEM

Volume/Issue : 2026;03(03)

Page No : 343-351

The rapid growth of the global startup ecosystem has created significant opportunities for innovation, economic development, and technological advancement. However, despite the increasing number of startups emerging each year, a large proportion of them fail during their early operational stages due to factors such as poor market fit, financial instability, inefficient resource management, and strategic misalignment. Identifying the underlying causes of startup failure is therefore essential for investors, founders, and policymakers to make informed decisions and reduce financial risk. This project, titled “Hidden Pattern Mining in Failed Business Startups,” aims to develop an intelligent analytical platform capable of predicting startup failure risk and uncovering hidden patterns associated with unsuccessful ventures. The proposed system, called the Startup Failure Risk Intelligence Platform (SFRIP), integrates full-stack web technologies with advanced machine learning techniques to analyze structured startup data such as funding history, industry type, team size, operational duration, and burn rate. The platform employs the XGBoost algorithm for predictive classification to estimate the probability of startup success or failure. In addition, K-Means clustering is used to segment startups into meaningful risk categories, while the FP-Growth algorithm is applied to discover frequent patterns and associations that commonly occur in failed startups. These analytical techniques enable the system to generate both quantitative risk predictions and qualitative insights into contributing factors. The results are presented through interactive dashboards that transform complex analytical outputs into intuitive visualizations for non-technical users. By combining predictive analytics, pattern mining, and visualization within a web-based system, the project demonstrates how data-driven intelligence can enhance investment decision-making, support startup founders in identifying weaknesses, and ultimately contribute to reducing failure rates within the entrepreneurial ecosystem.

Keywords :

Author : SK. AHMED MOHIDDIN, GOPISETTI BHANUVARSHA, ERADALA NAGA MAHA JAYA SRI, PAVANA SRI RAM BASWANI, KUMBHAM BABY SWARUPA

Title : HIDDEN PATTERN MINING IN FAILED BUSINESS STARTUPS

Volume/Issue : 2026;03(03)

Page No : 352-361

Academic performance prediction has become an important area of research in the education sector as institutions aim to improve student success and learning outcomes. Identifying academically weak students at an early stage allows educators to provide timely guidance, mentoring, and remedial support. Traditional evaluation methods mainly rely on examination results, attendance records, and teacher observations, which often fail to capture the complex factors that influence student performance. With the availability of educational datasets and advancements in data analytics, machine learning techniques provide a powerful approach for analyzing academic data and predicting future outcomes. This project presents a Student Performance Prediction System that uses regression-based machine learning models to analyze student academic data and predict final performance. The system considers various factors such as study hours, attendance percentage, internal assessment marks, and other academic indicators that influence student results. The collected dataset is preprocessed to handle missing values, normalize features, and prepare the data for model training. Machine learning algorithms such as Linear Regression and Random Forest Regression are applied to build predictive models. The models are trained and evaluated using performance metrics including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R² Score to measure prediction accuracy. Experimental results demonstrate that the Random Forest Regression model provides better prediction performance compared to Linear Regression due to its ability to capture complex relationships between variables. The proposed system helps educational institutions move from traditional evaluation methods to a data-driven approach, enabling early identification of at-risk students and supporting improved academic planning. Overall, this system provides an efficient and scalable solution for enhancing student performance analysis and decision-making in educational environments.

Keywords :

Author : MD.AHMED,KADIMI DEEPIKA,GUMMADI PAVANI NAGA LAKSHMI DURGA,CHANDIKA BALA NAGA VAMSI,ABDUL KHUDUS

Title : ACADEMIC PERFORMANCE ANALYSIS AND PREDICTION USING REGRESSION TECHNIQUES

Volume/Issue : 2026;03(03)

Page No : 362-371

The rapid growth of job applications in modern recruitment has made manual resume screening a time-consuming and inefficient process for human resource professionals. To address this challenge, the proposed system, Automated Resume Screening System Using Natural Language Processing (NLP), aims to streamline and improve the candidate selection process through intelligent automation. The system analyzes and evaluates resumes by extracting relevant information such as skills, qualifications, experience, and keywords related to job requirements. Using Natural Language Processing techniques and machine learning algorithms, the system processes unstructured resume data and converts it into structured information that can be easily compared with job descriptions. The model then ranks and filters candidates based on their relevance to the specified job profile, helping recruiters quickly identify the most suitable applicants. This approach reduces manual effort, minimizes human bias, and significantly improves the efficiency and accuracy of recruitment processes. The system typically involves stages such as resume data collection, text preprocessing, feature extraction, skill matching, and candidate ranking. By applying techniques such as tokenization, stop-word removal, and vectorization, the system understands the semantic meaning of resume content and evaluates the compatibility between candidate profiles and job requirements. The proposed solution can be integrated into recruitment platforms to assist organizations in handling large volumes of applications while ensuring fair and objective candidate evaluation. Overall, the system demonstrates how artificial intelligence and natural language processing can transform traditional hiring practices into a faster, more reliable, and data-driven process, ultimately helping organizations select qualified candidates efficiently.

Keywords :

Author : B. INDRA DEVI, BHATRAJU SAILU, KARRA AKHILA, MOHAMMAD MALLIKA, GANTASALA SARATH VENKATA NAGARAJU

Title : RESUME SCREENING SYSTEM USING NLP

Volume/Issue : 2026;03(03)

Page No : 372-381

The distributed computing and peer-to-peer cloud platforms boom have rendered it incredibly critical to create security systems capable of fulfilling authentication, privacy and key management simultaneously. All the current protocols divide these issues into distinct sections and this complicates configuration and creates security vulnerability. This paper presents a comprehensive Authentication and Key Agreement (AKA) model to be used in peer-to-peer cloud environments in which the identity of the user remains anonymous throughout the entire stream of communication. The system is based on elliptic-curve-cryptography to authenticate each other, an ECDH-based session key and message protection with an AES 256_GCM. A pseudonymous registration uses real identities but makes them invisible, and provides accountability by using tamper-resistant logs that are hashed. Role Based Access Control is used to have fine grained permissions and secret sharing by shamir allows a group to construct a shared key as many party is present. The cloud storage is protected such that all the files remain encrypted even when they are stored in the services of third parties. Our prototype has been developed using Flask and Python with SQLite database. Benchmarks depict that the structure is low overhead, replay resistant, impersonation resistant and simple to deploy through browser interface. The modular unit is, also, the fact that we can later replace primes with post-quantum primes or put it into alternative modes of operation.

Keywords : authenticity and key agreement, an anonymous identity, elliptic-curve cryptography, peer-to-peer cloud computing, secret sharing, Shamers principle and role-based access control.

Author : Mrs P Jayasree, Miryabbelli Manoj Kumar, Dogga Sai Jaswanth, Korra Sachin, Tedlapu Hema

Title : Authentication and Key Agreement Based on Anonymous Identity for Peerto-Peer Cloud

Volume/Issue : 2026;03(03)

Page No : 382-390

Records of land and property land are essentially the mainstay of any economy. Whenever there is a wrong or incomplete ownership records, it strikes people, courts and governments rather hard, in money and time. It has been estimated by World Bank that over 70 percent of the population resides in areas where land tenure is not well managed particularly in the developing countries. Although some attempts have been made towards digitization over the years, majority of the land registries continue to use centralized databases that have concentrated all the authority among an individual. That exposes them to hacking or in-house interferences which translate into loss of data.

Keywords : Blockchain, Ethereum, smart contracts, land registration, decentralized application, Solidity.

Author : Dr K N S Lakshmi,D Kiran, P Bhaskar Rao, K Mohit, D Vivek Raj, V Manikanta

Title : GREENLAND: A SECURE LAND REGISTRATION SCHEME FOR BLOCKCHAIN

Volume/Issue : 2026;03(03)

Page No : 391-394

Quantum Machine Learning (QML) has emerged as a compelling research domain that unites quantum computing principles with classical learning theory to address computationally demanding classification problems. This paper investigates two principal QML architectures—Variational Quantum Classifiers (VQC) and Quantum Neural Networks (QNN)—within the context of binary classification. We present an end-to-end hybrid quantum-classical pipeline applied to the UCI Mushroom dataset, comprising 22 categorical attributes, and demonstrate that quantumenhanced models can achieve competitive classification performance under present-day hardware constraints. The pipeline integrates Quantum Random Access Code (QRAC) encoding for dimensionality reduction, ZZFeatureMapbased quantum feature mapping, RealAmplitudes parameterized ansatz, and classical optimizer-driven parameter updates using the COBYLA algorithm. The complete system is deployed through a Flask web interface and containerized with Docker for reproducibility. Experimental results on a quantum simulator indicate a classification accuracy of 87.4%, precision of 0.891, and recall of 0.863, establishing the practical viability of NISQ-era quantum classifiers for structured categorical data. The paper further analyzes convergence behavior, circuit depth effects, and comparative performance against classical baselines, providing insights into the strengths and limitations of current QML architectures for real-world classification deployment.

Keywords : Quantum Machine Learning, Variational Quantum Classifier, Quantum Neural Network, QRAC Encoding, NISQ, Qiskit, Binary Classification

Author : Dr K N S Lakshmi, Bondalapu Sandhya, Nasari DurgaPrasad, Agraharapu DeviPrasad, Penugonda Rishitha

Title : Quantum Machine Learning for Classification Tasks:Variational Quantum Classifiers and Quantum Neural Networks

Volume/Issue : 2026;03(03)

Page No : 395-400

Modern distributed computing environments require secure authentication across organizational boundaries without exposing sensitive identity credentials. Existing cross-domain authentication solutions rely on centralized trusted third parties, creating single points of failure and privacy vulnerabilities. This paper proposes a Blockchain-Based Privacy Preserving Cross Domain Authentication Scheme (BP-CDAS) combining zero-knowledge proof (ZKP) protocols for privacy-preserving credential verification, Ethereum smart contracts for trustless authentication management, and decentralized identifiers (DIDs). The scheme eliminates centralized certificate authorities by distributing trust across a permissioned blockchain consortium. Security analysis demonstrates resistance to impersonation, replay, and man-inthe-middle attacks. Performance evaluation shows average authentication latency of 1.8 seconds with throughput of 180 authentications per minute, suitable for enterprise federated identity management.

Keywords :

Author : Mrs J.Sri Maha Lakshmi, P.Yaswanth kumar, S.Abhinash, S.Haripriya,T.Gowri Santhoshi

Title : Blockchain Based Privacy Preserving Cross Domain Authentication Scheme

Volume/Issue : 2026;03(03)

Page No : 401-406

Dog breed identification is a challenging fine-grained visual classification task due to high inter-breed similarity and substantial intra-breed variation in appearance, pose, and lighting. Accurate automated identification is valuable for pet management, veterinary care, lost pet recovery, and animal research. This paper proposes a Hybrid Deep Learning approach that combines the complementary strengths of ResNet-101, InceptionV3, and Xception architectures through ensemble feature fusion and weighted majority voting. The hybrid system leverages transfer learning from ImageNet pre-trained weights and applies extensive data augmentation to improve generalization on the Stanford Dogs dataset (120 breeds, 20,580 images). A comparative analysis evaluates individual and hybrid models. Experimental results demonstrate that the proposed hybrid approach achieves 91.4% top-1 accuracy, outperforming individual models by 3-7%, offering a robust solution for automated dog breed recognition.

Keywords : Dog Breed Identification, Hybrid Deep Learning, ResNet-101, InceptionV3, Xception, Transfer Learning, Fine-Grained Visual Classification, CNN Ensemble

Author : Mr. R. Bhanu Sankar, D. Vasanth, Yuvaraj Singh, T. Harish, Y. Kumar

Title : HYBRID DEEP LEARNING ALGORITHMS FOR DOG BREED IDENTIFICATION

Volume/Issue : 2026;03(03)

Page No : 407-411

Accurate and timely identification of flood-affected regions is critical for disaster response, emergency resource allocation, and damage assessment. This paper proposes a deep learning-based flood region segmentation system using an enhanced SegNet encoder-decoder architecture adapted for multi-source remote sensing image analysis. The proposed model is augmented with skip connections, attention gates, and multi-scale input processing to improve flood boundary delineation accuracy. The system achieves a mean IoU of 87.3% for flood region segmentation on the Copernicus Emergency Management Service dataset, outperforming baseline SegNet, U-Net, and FCN models. The framework produces pixel-wise flood maps within seconds, supporting rapid disaster response workflows with strong generalization across diverse geographic environments and flood event types.

Keywords :

Author : Mr. G. Ravi Kumar,S. Pavan Kalyan, S. L. Sri Harshitha, P. Sai Charan, CH. Jnana Sowmya

Title : Flood Region Segmentation Using SegNet Deep Neural Networks

Volume/Issue : 2026;03(03)

Page No : 412-417

Accurate segmentation of retinal blood vessels from fundus images is essential for early detection of diseases such as diabetic retinopathy, glaucoma, and hypertension. Manual analysis is time-consuming, subjective, and dependent on expert availability. This paper presents an automated retinal blood vessel segmentation system using the U-Net encoder-decoder architecture trained on the DRIVE and STARE datasets. The proposed system applies CLAHE contrast enhancement, normalization, and augmentation as preprocessing steps followed by U-Net with skip connections for precise vessel boundary preservation. The model is evaluated using pixel accuracy, sensitivity, specificity, Dice coefficient, and AUC-ROC. Experimental results demonstrate superior segmentation performance compared to traditional image processing and prior deep learning baselines, achieving 95.8% accuracy and 0.823 Dice coefficient on the DRIVE dataset.

Keywords : U-Net, Retinal Blood Vessel Segmentation, Fundus Image, Diabetic Retinopathy, Deep Learning, DRIVE, STARE, Medical Image Analysis

Author : Mr. G. Ravi Kumar, M. Kalyan, D. Praveena, K. Yugandhar, P. Deepika

Title : U-NET BASED RETINAL BLOOD VESSEL SEGMENTATION FOR EARLY DISEASE DETECTION

Volume/Issue : 2026;03(03)

Page No : 418-423

The detection of surface defects in solar cells is crucial for ensuring the efficiency and reliability of photovoltaic systems. This paper proposes an automated solar cell surface defect detection system based on an optimized YOLOv5 deep learning model. The system detects and localizes multiple defect types including micro-cracks, scratches, contamination, broken grids, and discoloration in real time. Optimization strategies including custom anchor tuning, mosaic augmentation, and transfer learning maximize detection performance. Experimental results demonstrate mAP@0.5 of 94.3%, precision of 93.1%, recall of 91.8%, with real-time inference at 48 FPS.

Keywords :

Author : Mrs. P.G.V. Rekha, B.Priyadarshini, B.Lavanya, N.Sai ManiKumar, K.Manmohan Singh

Title : Solar Cell Surface Defect Detection Based on Optimized YOLOv5

Volume/Issue : 2026;03(03)

Page No : 424-428

The explosive growth of digital visual data has created a pressing need for accurate and efficient image retrieval systems. This paper proposes a Content-Based Image Retrieval (CBIR) system using a hybrid feature fusion framework combining deep learning-based and handcrafted visual features. Low-resolution images are first enhanced using SRCNN super-resolution. CNN features are extracted using ResNet50 and VGG16, while handcrafted features include color histograms, LBP, and HOG. Experiments on Corel-1K demonstrate mAP of 87.4% and top-5 accuracy of 91.2%, outperforming single-feature baselines.

Keywords :

Author : Mr.K.Gurucharan, G.BharatChand, G.VijayendraVarma, CH.OmVardhan, B.RohithKumar

Title : Content-Based Image Retrieval for Super-Resolution Images Using Feature Fusion: Deep Learning and Handcrafted Features

Volume/Issue : 2026;03(03)

Page No : 429-433

Centralized photo sharing platforms such as social media and cloud storage services face critical challenges including data privacy violations, unauthorized access, content tampering, and loss of user ownership control. This paper proposes a Blockchain-Based Secure Photo Sharing System that ensures decentralized, transparent, and tamper-proof digital photo management. Photos are stored in the InterPlanetary File System (IPFS) for efficient decentralized storage, while cryptographic hashes, ownership records, and access permissions are stored on the Ethereum blockchain. Solidity smart contracts automate user authentication, access control, and permission management without reliance on third parties. The system guarantees data integrity through hash-based tamper detection, enabling users to maintain full ownership and securely grant or revoke sharing rights. Experimental results demonstrate 100% tamper detection, sub-second access verification, and elimination of single-point-offailure vulnerabilities present in centralized systems.

Keywords : Blockchain, IPFS, Photo Sharing, Smart Contracts, Decentralization, Data Integrity, Cryptographic Hash, Access Control, Privacy

Author : Mr. D. Pramod Kumar, D.Sheeja, P.Santhoshi, G.Sai, V.Anil

Title : SECURE PHOTO SHARING ACROSS PLATFORMS USING BLOCKCHAIN TECHNOLOGY

Volume/Issue : 2026;03(03)

Page No : 434-438

The study entitled “A Study on Fund Flow Statement of Muthoot Fincorp Limited” analyzes the movement of funds for a period of five years (2020–2024). The Fund Flow Statement helps in understanding how funds are generated and utilized between two balance sheet dates. The study is based on secondary data collected from annual reports and RBI publications. The analysis highlights the sources and applications of funds, changes in working capital, and the financial strategy adopted by the company.

Keywords : Fund Flow Statement, Working Capital, Sources of Funds, Application of Funds, NBFC, Financial Strategy.

Author : T.Likhitha,B.Nagaveni

Title : A study on Funds flow Statement 0n Muthoot Fincorp ,Jammalamadugu

Volume/Issue : 2026;03(03)

Page No : 517-523