Journal of Science Engineering Technology and Management Science

Plant diseases pose a significant threat to agricultural productivity and food security, making early and accurate detection essential. Traditional image processing and machine learning techniques often struggle to handle variations in lighting, background noise, and complex disease patterns. To address these challenges, this paper proposes a hybrid plant disease prediction system that integrates quantum-inspired image processing with deep neural networks. The proposed approach utilizes angle-based quantum encoding to transform leaf images into quantum-like representations, enabling efficient and discriminative feature extraction. Statistical measurements derived from these representations are combined with color and texture features to form a robust feature set. These features are then fed into a deep neural network model for multi-class disease classification. The system is designed to be scalable and robust, capable of handling diverse plant disease categories under varying environmental conditions. Experimental results demonstrate that the proposed method significantly improves classification accuracy and reliability compared to conventional approaches. The developed framework can be effectively used in real-world agricultural decision support systems for early disease diagnosis and crop management.

Keywords : Plant Disease Detection, Quantum-Inspired Image Processing, Deep Neural Networks (DNN), Feature Extraction, Image Classification, Agricultural AI, Leaf Image Analysis, Machine Learning, Pattern Recognition, Smart Farming

Author : T.S.R.Krishna Prasad,Kollati Ribka,J Venkata Guru Prasad,Jyothi Ayyappa Prasanna Kumar,Madugula Ajay Kumar

Title : Plant disease detection using quantum image processing

Volume/Issue : 2026;03(04)

Page No : 688-697

Chronic Heart Failure (CHF) is a serious cardiovascular condition associated with high morbidity and mortality worldwide. Early and accurate detection of CHF can significantly enhance patient outcomes and reduce healthcare costs. Traditional diagnostic methods, including echocardiography and clinical assessment, are often expensive and require specialized expertise. This project proposes a machine learning-based system to classify CHF using heart pulsation data derived from electrocardiograms (ECG) and photoplethysmography (PPG) signals. Signal preprocessing enhances data quality by removing noise and artifacts. Time-domain and frequency-domain features are extracted to characterize pulsation patterns. Multiple machine learning classifiers such as Support Vector Machines (SVM), Random Forests, and Artificial Neural Networks (ANN) are trained on labeled datasets. Feature selection reduces dimensionality and improves model performance. The system identifies CHF with high accuracy, precision, recall, and F1-score. Cross-validation ensures model robustness. The model supports real-time prediction capabilities for wearable devices. Visualization dashboards provide insight into detected patterns. Automated alerts notify healthcare professionals of suspected CHF cases. The approach reduces dependence on costly clinical diagnostics. The system contributes to accessible and scalable CHF screening. Ethical data handling and privacy compliance are maintained.

Keywords : Chronic Heart Failure (CHF) Machine Learning Classification Heart Pulsation Signals ECG/PPG Feature Extraction

Author : 1Kolli Sivalakshmi, 2Pothireddy Susanna, 3Pujari Somnath, 4Sadhu Deepthi, 5Ravi Mani Deepthi, 6Mrs. L Mounika,

Title : Utilizing Machine Learning for the Identification of Chronic Heart Failure (CHF) from Heart Pulsations

Volume/Issue : 2026;03(04)

Page No : 698-703

Accurate discrimination between natural earthquakes and artificial explosions is critical for seismic monitoring, disaster management, and national security. Traditional seismic analysis techniques rely heavily on handcrafted features and expert interpretation, which can be time-consuming and error-prone. With the advancement of deep learning, automated seismic signal classification has gained significant attention. This project proposes a Deep Convolutional Neural Network (DCNN)-based approach to classify seismic events as earthquakes or explosions. Seismic waveform data are transformed into time-frequency representations such as spectrograms for effective feature learning. The CNN model automatically extracts discriminative spatial and temporal features from seismic signals. Data preprocessing enhances signal quality and reduces noise interference. The trained model achieves high classification accuracy without manual feature engineering. Performance is evaluated using accuracy, precision, recall, and F1-score metrics. The system supports rapid and reliable seismic event classification. Explainability tools provide insight into model decisions. The proposed framework improves detection reliability in real-time seismic monitoring systems. This approach enhances both scientific research and security surveillance. Overall, the system offers a scalable and intelligent solution for seismic event classification.

Keywords : Seismic Signal Classification Deep Convolutional Neural Networks (DCNN) Earthquake Detection Explosion Identification Time-Frequency Analysis

Author : 1Pulivarthi Raghuvardhan, 2Karrothu Chaitanya Kumar, 3Gummadi Praveen, 4Y Srinu, 5Mrs.B Roja Lakshmi

Title : Using Deep Convolutional Neural Networks for Earthquake and Explosion Classification

Volume/Issue : 2026;03(04)

Page No : 704-708

Secure file sharing is a critical requirement in modern digital ecosystems, especially with increasing concerns about data privacy, integrity, and unauthorized access. Traditional centralized storage systems often suffer from single-point failures, data tampering, and privacy vulnerabilities. This project proposes a decentralized secure file sharing system leveraging Blockchain and InterPlanetary File System (IPFS). Blockchain ensures immutability, transparency, and tamper-evident audit trails for file transactions. IPFS provides distributed, content-addressed storage that eliminates reliance on centralized servers. Users can upload, share, and retrieve files securely using cryptographic hashes. Smart contracts govern access control, permission updates, and file metadata management on the blockchain. File content itself is stored on IPFS, while its reference hash is recorded onchain for verification. End-to-end encryption ensures confidentiality of sensitive data. Consensus mechanisms prevent unauthorized modifications. The system supports user authentication and role-based access control. Decentralized storage enhances fault tolerance and availability. Real-time tracking of file changes prevents data tampering. Auditable logs enable forensic investigations. Scalability is achieved through distributed networks. The architecture supports interoperability with existing storage solutions. This approach strengthens data security, privacy, and trust for collaborative environments. Overall, blockchain and IPFS jointly provide a scalable and secure file sharing framework for modern applications.

Keywords : Secure File Sharing Blockchain Technology Smart Contracts IPFS (InterPlanetary File System) Decentralized Storage

Author : 1Majji Pavan Kumar, 2Rebba Surya, 3Gogula Ajay, 4Sake Sravanthi, 5Mr.V.V.B. Chari

Title : Secure File Sharing with Blockchain And IPFS

Volume/Issue : 2026;03(04)

Page No : 709-715

The Online Voting System using Blockchain is an innovative approach aimed at enhancing the transparency, security, and reliability of electoral processes. Traditional voting systems often suffer from issues such as vote manipulation, lack of transparency, and centralized control. This project introduces a decentralized voting mechanism using blockchain technology to eliminate these challenges. Blockchain ensures that all votes are recorded in a secure, immutable ledger, making it nearly impossible to alter or tamper with voting data. Each vote is encrypted and stored as a block, ensuring voter anonymity and data integrity. Smart contracts are utilized to automate vote validation and counting processes, reducing human intervention and errors. The system provides real-time access to election results while maintaining strict security protocols. Voters can participate remotely using secure authentication methods, increasing accessibility and voter turnout. The decentralized nature of blockchain removes the dependency on a central authority, enhancing trust among users. Additionally, the system prevents double voting and unauthorized access through cryptographic techniques. The implementation of this system can significantly improve election efficiency and fairness. It is scalable and adaptable for various types of elections, including governmental and organizational voting. Overall, this solution aims to modernize voting systems by leveraging cutting-edge technology. The proposed system ensures transparency, accountability, and accuracy

Keywords :

Author : 1B. Naga Jagadish, 2B. Rajani Kumari, 3D. Naga Ratnam, 4R. Shanmukha Rao, 5Mr. G Koteswara Rao

Title : ONLINE VOTING SYSTEM USING BLOCK CHAIN

Volume/Issue : 2026;03(04)

Page No : 716-723

Hypertension is one of the most common chronic health conditions and a major risk factor for cardiovascular diseases worldwide. Early detection and continuous monitoring are essential to prevent severe complications such as stroke, heart attack, and kidney failure. Traditional diagnosis relies on periodic blood pressure measurements and clinical assessment, which may miss early warning signs. This project proposes a machine learning-based healthcare system for the prediction of hypertension using patient health data. The system analyzes clinical parameters such as age, body mass index (BMI), blood pressure readings, cholesterol levels, lifestyle factors, and family history. Data preprocessing ensures quality and consistency of medical records. Multiple machine learning algorithms are trained to classify individuals as hypertensive or non-hypertensive. Feature selection improves model accuracy and efficiency. The system provides early risk prediction and decision support for healthcare professionals. Performance is evaluated using accuracy, precision, recall, and F1-score. Automated alerts notify users of potential hypertension risks. The model supports preventive healthcare and personalized treatment planning. Secure data handling ensures patient privacy. The proposed approach enhances accessibility, accuracy, and efficiency in hypertension management.

Keywords : Hypertension Prediction Machine Learning in Healthcare Predictive Analytics Chronic Disease Monitoring Health Risk Assessment

Author : 1Shaik Mubhashira, 2Rasipogula Swapna, 3Shaik Sadhik, 4T V Sai Charan,5Gunja Narendra, 6Mr.V.V.B. Chari

Title : Prediction of Hypertension & Healthcare System using ML

Volume/Issue : 2026;03(04)

Page No : 724-729

The rapid growth of online payment systems has significantly increased the risk of financial fraud. Fraudulent transactions cause substantial financial losses to individuals, banks, and e-commerce platforms. A major challenge in fraud detection is the highly imbalanced nature of transaction datasets, where fraudulent cases are rare compared to legitimate ones. Traditional machine learning models tend to be biased toward majority classes, resulting in poor fraud detection rates. This project proposes an Online Fraud Payment Detection system using balanced machine learning algorithms to address class imbalance. Data balancing techniques such as SMOTE and undersampling are applied to improve model performance. Multiple machine learning classifiers are trained and evaluated on balanced datasets. Features related to transaction amount, location, time, and user behavior are analyzed. The system accurately identifies fraudulent transactions in real time. Performance is evaluated using precision, recall, F1-score, and ROC-AUC metrics. The model minimizes false negatives, which are critical in fraud detection. Automated alerts notify stakeholders of suspicious activities. The system improves decision-making for financial institutions. Scalability ensures handling of high-volume transactions. Security and data privacy are maintained throughout the process. The proposed approach enhances fraud detection accuracy and reliability. Overall, the system provides a robust and efficient solution for online payment fraud prevention.

Keywords : Online Payment Fraud Machine Learning Class Imbalance SMOTE Fraud Detection Systems

Author : 1Shaik Siddhik, 2Shaik Mohith, 3Gunja Sirisha, 4Yadagiri Vasantha Laxmi, 5Mr.N. Anjaneyulu

Title : Online Fraud Payment Detection using Balanced ML Algorithms

Volume/Issue : 2026;03(04)

Page No : 730-735

In recent years, the growing urgency to address environmental challenges has brought green finance to the forefront of academic and policy discussions. Against this backdrop, the present study examines how financial considerations and sustainability imperatives are shaping investment behaviour in the renewable energy sector. The paper seeks to understand not only the motivations driving investors and firms toward green investments but also the practical constraints that continue to influence their decisions. Drawing on both primary insights and secondary evidence, the study adopts an analytical approach to explore the role of environmental awareness, regulatory support, and financial incentives in guiding investment choices. The findings suggest that while there is a clear and increasing shift toward sustainability-oriented investments, decision-making remains strongly influenced by perceived risk, cost structures, and policy stability. Investors appear more inclined toward renewable energy projects when supported by transparent regulations, fiscal incentives, and long-term viability signals. At the same time, the study highlights those barriers such as high initial capital requirements, information asymmetry, and uncertainty in policy implementation continue to limit the full potential of green finance. These challenges underline the need for a more integrated framework where financial markets, regulatory bodies, and corporate strategies align toward sustainability goals. The paper concludes that strengthening green finance mechanisms and improving investor confidence are critical to accelerating the transition toward a sustainable energy future. It offers practical implications for policymakers and financial institutions to design more effective instruments that can bridge the gap between sustainability ambitions and investment realities.

Keywords : Green Finance, Corporate Sustainability, Renewable Energy, Investment Behaviour, Sustainable Investment

Author : P. Shirisha, Kavali Rudransh, Vadlakonda Kavitha

Title : GREEN FINANCE AND CORPORATE SUSTAINABILITY: INVESTMENT BEHAVIOUR IN THE RENEWABLE ENERGY SECTOR

Volume/Issue : 2026;03(04)

Page No : 736-745

The rapid growth of startup ecosystems has created a strong need for reliable methods to evaluate the potential success of newly established companies. Traditional investment decisions often rely on subjective judgment, limited historical analysis, and manual evaluation, which may lead to inaccurate predictions and high financial risk. This research proposes an efficient and novel machine learning–based approach for predicting startup company success rates using advanced ML paradigms. The proposed system integrates multiple data sources such as financial metrics, founder experience, market trends, funding history, product innovation, and customer engagement indicators to build a comprehensive predictive model. The methodology involves data preprocessing, feature engineering, and the application of supervised learning algorithms including Random Forest, Support Vector Machine, Gradient Boosting, and Neural Networks. Ensemble learning techniques are employed to improve prediction accuracy and reduce model bias. The system utilizes historical startup datasets to train and validate models, enabling identification of critical success factors influencing startup growth and sustainability. Performance evaluation is conducted using accuracy, precision, recall, and F1-score metrics to ensure robust model performance. Experimental results demonstrate that the proposed hybrid ML framework significantly improves prediction reliability compared to traditional statistical methods. The model assists investors, entrepreneurs, and policymakers in making datadriven decisions by providing early insights into startup viability. Overall, the proposed approach enhances risk assessment, optimizes investment strategies, and contributes to the development of intelligent decision-support systems within entrepreneurial ecosystems.

Keywords :

Author : Mr. L. N. V. Rao1 , KONDETI ANAND PAL2 ,RAVULAKOLLU LAKSHMI NARASAMMA3 ,METHUKUMILLI DIVYA4 ,KUNUKU SRI LAKSHMI5

Title : TRANSFORMING HAND GESTURE INTO WORDS A NOVEL APPROCH FOR ASSISTIVE COMMUNICATIONS USING ML TECHNIQUES

Volume/Issue : 2026;03(04)

Page No : 746-754

The rapid growth of online payment systems and digital transactions has significantly increased the risk of financial fraud, making fraud detection an essential component of modern financial security. Online fraud payment detection aims to identify suspicious or unauthorized transactions in real time to prevent financial losses for both users and financial institutions. However, one of the major challenges in fraud detection is the imbalance in transaction datasets, where fraudulent transactions represent only a small portion compared to legitimate ones. Traditional machine learning models often fail to accurately detect fraud due to this class imbalance problem. This study proposes an online fraud payment detection system using balanced machine learning algorithms to improve detection accuracy and reliability. The proposed approach applies data balancing techniques such as oversampling, under sampling, and synthetic data generation to ensure equal representation of fraudulent and non-fraudulent transactions during model training. Balanced algorithms including Random Forest, Support Vector Machine, Logistic Regression, and Gradient Boosting are utilized to enhance classification performance. Feature engineering and preprocessing techniques are also employed to extract meaningful transaction patterns and reduce noise in the dataset. The system evaluates performance using metrics such as accuracy, precision, recall, F1-score, and ROC-AUC, which are more suitable for imbalanced datasets. Experimental results demonstrate that balanced machine learning models significantly improve fraud detection rates while minimizing false positives. The proposed framework provides a scalable and efficient solution for secure online payment environments, enabling financial organizations to detect fraudulent activities proactively and enhance customer trust.

Keywords :

Author : Mr.V. SUDHAKAR1 , MORAMPUDI JOSHNITHA2 ,PILLI NAGA VIJAYA LAKSHMI3 , MUDDUNENI KOMALI NAGA SRI SAI KEERTIKA4 , MADDALI ROHITH BABU5

Title : ONLINE FRAUD PAYMENT DETECTIONUSING BALANCED ML ALGORITHMS

Volume/Issue : 2026;03(04)

Page No : 755-763

Stock price prediction has become an important research area in financial analytics due to the increasing availability of historical market data and advancements in machine learning techniques. The objective of this project is to develop an intelligent system that analyzes past stock market trends and predicts future price movements with improved accuracy. The proposed approach utilizes data preprocessing, feature engineering, and predictive modeling algorithms such as Linear Regression, Random Forest, and Long Short-Term Memory (LSTM) networks to capture both statistical patterns and temporal dependencies in stock data. Historical price data, including open, high, low, close, and trading volume, are used as input features to train the model. The system aims to assist investors and financial analysts in making informed decisions by providing trend forecasts and risk insights. Performance evaluation is carried out using metrics such as Mean Squared Error (MSE) and accuracy comparisons across different models. The results demonstrate that machine learning–based methods can effectively identify hidden patterns in financial markets, offering a scalable and data-driven solution for stock price prediction while reducing manual analysis effort.

Keywords :

Author : Venkata Ratnam Ganji1 , A. Jyotsna Sri2 , J.Lavanya3 , K. Hima Sankar4 , G.Navya5

Title : STOCK PRICE PREDICTION USING MACHINE LEARNING TECHNIQUES

Volume/Issue : 2026;03(04)

Page No : 764-774

Effective communication is a fundamental challenge for individuals with speech or hearing impairments. Traditional communication aids often require manual input or are limited in scope. This project proposes a novel machine learningbased approach to transform hand gestures into textual words, enabling real-time assistive communication. The system uses computer vision techniques to capture hand gestures and employs machine learning algorithms, including Convolutional Neural Networks (CNNs) and LSTM networks, to recognize and translate gestures into corresponding words. The approach focuses on accuracy, speed, and adaptability, allowing users to communicate efficiently without prior extensive training. Experimental results demonstrate that the system can accurately interpret diverse hand gestures and convert them into readable text, providing a practical and user-friendly assistive tool for individuals with communication difficulties.

Keywords :

Author : Mr. L.N.V. Rao1 , UMMADISETTI YASASRI2 , PARASA NARESH 3 , MERUGUMALA RAVI KIRAN4 , THOTA THARUN PRUDHVI5

Title : TRANSFORMING HAND GESTURE INTO WORDS A NOVEL APPROCH FOR ASSISTIVE COMMUNICATIONS USING ML TECHNIQUES

Volume/Issue : 2026;03(04)

Page No : 775-784

The increasing diversity of video games and the rapid expansion of digital gaming platforms have created a critical need for intelligent systems capable of accurately classifying video games into genres across multiple domains. Traditionally, genre classification relied on manual tagging or simple keyword-based approaches, and early machine learning methods used basic text representations such as Bag-of-Words or TF-IDF, which often failed to capture semantic meaning and struggled with crossdomain generalization. These limitations resulted in inconsistent accuracy, high manual workload, and limited adaptability for large or new datasets. To address these challenges, this study presents an automated, GUI-based machine learning framework that analyzes game descriptions and metadata to classify games into genres while also evaluating additional attributes such as graphics quality, soundtrack quality, and story quality. The system leverages Masked and Permuted Pre-training Network (MPNet) Transformer embeddings to extract deep semantic features, enabling robust cross-domain generalization. Multiple classification algorithms are integrated, including Ridge Classifier (RC), Nearest Centroid (NC), Restricted Boltzmann Machine - Ridge (RBMR) pipeline classifier, and a novel Ensemble of Oblique Trees (EOT) classifier. Experiments on balanced datasets generated via Random Under Sampling (RUS) demonstrate strong performance, with several models achieving high accuracy. The framework provides comprehensive evaluation metrics including precision, recall, F1-score, confusion matrices, and ROC values facilitating thorough model comparison and informed decisionmaking.

Keywords : Video game genre classification, digital gaming platforms, text analysis, semantic feature extraction, transformer embeddings, cross-domain generalization, metadata analysis, automated classification system

Author : P. Kamaraja Pandian, Rasala Sindhuja, Kommu Ruthvika, Kudikala ‘ushasri,

Title : MPNet Transformer Embeddings with Oblique Tree Ensemble for CrossDomain Video Game Genre and Quality Prediction

Volume/Issue : 2026;03(04)

Page No : 785-795

Mineral hardness and density are important physical properties used in geology, mining, and material science for mineral identification and classification. Traditionally, mineral hardness was determined using the Mohs hardness scale through manual scratch testing, where minerals were compared against standard reference minerals. This traditional approach required laboratory testing, expert knowledge, and significant time to obtain accurate results. In many cases, the process was labor-intensive and sometimes produced inconsistent outcomes due to human errors and variations in experimental conditions. To address these limitations, the proposed system introduces a machine learning based multimodel regression approach for forecasting mineral hardness and specific gravity using chemical and physical features of minerals. The system is implemented as a web-based application using the Django framework. Multiple regression algorithms are applied to improve prediction performance, including CatBoost Regressor (CBR), AdaBoost Regressor (ABR), Random Forest Regressor (RFR), and Greedy Tab Transformer (GTT). These models analyze the relationships between mineral properties and target variables to accurately predict Mohs hardness and specific gravity. Experimental evaluation shows that the GTT model outperforms the other models in terms of prediction performance. The system also provides functionalities such as model training, algorithm comparison, and prediction modules for analyzing mineral properties. By integrating machine learning models with an interactive web platform, the proposed system offers a faster, more efficient, and scalable method for predicting mineral hardness compared to traditional manual testing methods.

Keywords : Mineral hardness, Specific gravity, Mineral identification, Mohs hardness scale, Physical properties, Chemical composition, Mineral classification, Django framework.

Author : Ch. Mounika, kadem Vamshi Krishna, Gajjala Ravinandhana, Ponnam Bharath Kumar

Title : Mineral Hardness and Specific Gravity Estimation Using TabTransformer Embeddings for Geological Analysis

Volume/Issue : 2026;03(04)

Page No : 796-805

Real-estate valuation is essential for buyers, sellers, and investors, yet traditional approaches depend heavily on manual inspection, subjective judgment, and limited comparable data. Existing automated methods often frame valuation solely as a regression problem, overlooking classification aspects and failing to address imbalanced datasets. Additionally, conventional models struggle with non-linear relationships and high-dimensional features, leading to poor generalization and inconsistent pricing. This work proposes an integrated machine learning framework that combines preprocessing, Synthetic Minority Over-sampling Technique (SMOTE)–based class balancing, exploratory data analysis, and multi-model training for both regression and classification tasks. The regression module predicts key numerical attributes such as property price (in lakhs) and built-up area (in square feet), ensuring accurate continuous-value estimation across diverse property types. The classification module identifies property orientation into facing categories—North (N), South (S), East (E), and West(W)—important for design, ventilation, sunlight exposure, and valuation. The framework employs three model families: Gradient Boosting Classification and Regression Trees (Gradient Boosting CART), Extreme Gradient Boosting CART (XGBoost-CART), and a hybrid Long Short-Term Memory–TreeNet (LSTM–TreeNet) architecture. The hybrid model captures deep temporal and feature-level patterns from structured and unstructured data, followed by refinement through CART-based decision structures. Results show improved prediction accuracy, enhanced classification precision, reduced human bias, and greater consistency in valuation. The proposed system offers a scalable, data-driven solution for reliable and automated real-estate analytics.

Keywords : Real-estate valuation, property pricing, data preprocessing, SMOTE, exploratory data analysis, regression analysis, classification, imbalanced data, feature engineering, automated valuation system.

Author : B. Ramesh, Addetla Kavya, Bonam Udayasree, J Praveen Kumar

Title : Real Estate Predictive Analytics using Multi-Model Regression and Classification Pipeline with Recurrent Learning

Volume/Issue : 2026;03(04)

Page No : 806-818

Customer churn prediction is a critical analytical capability for organizations seeking to improve customer retention, optimize marketing strategies, and sustain revenue growth in competitive environments. With the expansion of digital interactions, businesses require intelligent systems to identify customers likely to discontinue services or reduce engagement. Traditional churn analysis relied on manual evaluation using basic statistical methods such as mean comparisons, threshold-based rules, and spreadsheet-driven reporting, which limited scalability and predictive accuracy. To address these limitations, this study proposes a Classification and Regression Tree (CART) based predictive framework that applies supervised machine learning to analyze multiple customer attributes, including demographics, purchase history, and digital behavior, for estimating churn risk and customer engagement. The system evaluates four machine learning models: Passive-Aggressive (PA), Support Vector Machine (SVM), Extra Trees (ET), and Poly-BoostNet. The PA model enables fast incremental learning but shows lower accuracy due to limited nonlinear modeling capability. The SVM model improves prediction stability but struggles with complex feature interactions. The ET model enhances ensemble learning performance but still lacks optimal precision. The Poly-BoostNet model integrates Categorical Boosted Learning with Recurrent Polynomial Network (RPN) feature expansion, resulting in improved feature representation and more accurate decision boundaries. Experimental results demonstrate that Poly-BoostNet outperforms other models, achieving the highest R²-score for customer satisfaction prediction and 100% accuracy in Engagement with Advertisements classification. The system is implemented using machine learning techniques and Flask for web-based deployment, delivering a scalable, end-to-end pipeline for data preprocessing, model training, evaluation, and realtime prediction.

Keywords : Customer Churn Prediction, Customer Retention, Predictive Analytics, Customer Engagement, Data Analytics, Digital Behavior Analysis, Demographic Analysis, Purchase History Analysis, Supervised Learning Framework.

Author : P. Vyshali, Cheruku Pranathi Goud, Yala Bhargavi, Kodam Bharani Kumar, Karingula Ranga Reddy

Title : Poly-BoostNet Enhanced Consumer Churn Analytics With CART-Driven Insights for Business Decision Systems

Volume/Issue : 2026;03(04)

Page No : 819-830

Modern civilization is seeing an increase in the popularity of online voting. It has a tremendous chance of bringing down administrative expenses and raising voting turnout. Voting may be done everywhere there is an Internet connection, thus there is no need to waste a lot of money in the aspect of the setting up of voting places for people to vote in. Online voting options are considered with great suspicion despite these advantages because they pose new security risks. Vote has been rigged a lot of times on a grand scale and may result from a single vulnerability. Elections must use lawful, accurate, safe, and convenient electronic voting technology. But potential issues with electronic voting methods might prevent widespread implementation. These problems were addressed by the development of blockchain technology, which provides decentralised blockchain nodes for electronic voting and is utilised to construct digital voting.

Keywords :

Author : 1 Dr.P.EPSIBA, 2 KOLTUR RISHI, 3 ADURLA DIKSHITH REDDY, 4 GOLLA PAVAN KUMAR YADAV, 5 VISHWANADULA SAKETH

Title : AN EFFICIENT VOTE COUNTING SYSTEM WITH AADHAR VERIFICATION THROUGH BLOCK CHAIN

Volume/Issue : 2026;03(04)

Page No : 831-838

In many countries, traffic congestion has become a major challenge, making effective traffic management increasingly difficult for authorities. Issues such as optimal route selection based on congestion levels, timely communication with users, and real-time allocation of parking spaces add to the complexity. Although several approaches have been proposed for predicting parking prices, many existing solutions tend to favor either users or parking owners and often lack adequate security measures To overcome these limitations, this paper introduces a blockchain and Interplanetary File System (IPFS)-based parking price prediction framework, known as PAADAV. The proposed system enables users to reserve parking spaces in a secure and efficient manner while ensuring transparency, privacy, and trust for both users and parking space providers. The integration of blockchain technology ensures data integrity and security, while IPFS supports decentralized and cost-effective data storage. Additionally, the system incorporates a secondprice auction mechanism to determine optimal parking prices, benefiting both users and parking space owners. To evaluate its performance, the PAADAV model is simulated with 100 users and 40 parking slots under different auction strategies. Key performance metrics analyzed include user profit, owner profit, overall system revenue, scalability, computation time, and storage cost. The results demonstrate that the proposed framework not only enhances security and transparency but also provides economic benefits to both users and parking slot owners. Overall, PAADAV offers an efficient and reliable solution for smart parking management in modern urban environments.

Keywords :

Author : 1 Mrs.P.JANAKI, 2 PALLE PRABHAS, 3 B PRANEETH REDDY, 4 ETTINI DEVENDHAR, 5 BONDALA SRIVANI

Title : PAADAV-SMART PARKING BLOCKCHAIN AND IPFS

Volume/Issue : 2026;03(04)

Page No : 839-852

Water scarcity has become a critical global issue due to rapid population growth, climate change, and inefficient water management practices. Many regions face challenges in accessing, distributing, and conserving water resources effectively. This project proposes a centralized knowledge-sharing platform to address water scarcity by integrating data, insights, and best practices from multiple stakeholders such as government agencies, researchers, and local communities. The platform aims to facilitate efficient water resource management through real-time information sharing and data-driven decisionmaking. The system collects and analyzes data from various sources, including rainfall patterns, water usage statistics, reservoir levels, and environmental conditions. Advanced technologies such as data analytics and machine learning are utilized to identify trends, predict water demand, and recommend optimal resource allocation strategies. The platform enables users to share knowledge, report issues, and access guidelines for sustainable water usage. Additionally, it supports collaborative decision-making by providing a unified interface for monitoring and managing water resources. Experimental results indicate that the centralized approach improves water distribution efficiency, reduces wastage, and enhances awareness among users. The system also helps in early detection of water shortages and supports proactive planning. However, challenges such as data integration, scalability, and user adoption need to be addressed. Overall, this project highlights the importance of leveraging technology and collaboration to tackle water scarcity and promote sustainable water management practices.

Keywords :

Author : 1 Mrs.C.MADHUSUDANAMMA, 2 M SNEHA , 3 JEEDI MANVITHA , 4 VEERAMALLA MANIDEEP

Title : WATER SCARITY MANAGEMENT THROUGH CETRALIZED KNOWLEDGE SHARING PLATFORM

Volume/Issue : 2026;03(04)

Page No : 853-859

Phishing is a type of cyberattack in which attackers design fraudulent websites that closely imitate legitimate platforms to deceive users into revealing sensitive information such as login credentials, banking details, or personal data. Recent studies highlight the effectiveness of machine learning (ML) techniques in detecting phishing websites by analyzing features such as URL structures, webpage content, and domain-related information. Although these methods achieve high accuracy, they often lack real-time protection and seamless integration into user-friendly applications. To overcome these limitations, this project proposes a browser extension powered by a machine learning model that can detect phishing attempts during web browsing. The system extracts important features from web pages and processes them through a lightweight yet efficient ML classifier, which provides instant alerts to users when a potential threat is identified. The model is trained on a balanced dataset containing both legitimate and phishing websites to ensure reliable and consistent performance. By combining real-time detection with an easy-touse interface, the proposed solution enhances user safety and provides an effective defense mechanism against phishing attacks in everyday internet usage.

Keywords : Machine learning, Phishing, Website detection

Author : 1 Mrs.G.PRIYANKA,2 VANGALA BHANU , 3 BASU NAVEEN KUMAR , 4 DANDUGULA PRASHANTH , 5 JUKANTI SATHVIKA

Title : PHISHING WEBSITE DETECTION USING MACHINE LEARNING

Volume/Issue : 2026;03(04)

Page No : 860-868

Modern financial systems still rely on manual coin sorting and counting, which are both laborintensive and susceptible to human error. Existing automation approaches, typically based on weight sensors or simple image processing methods such as edge detection and template matching, tend to perform inadequately in practical scenarios. This is largely due to challenges like varying lighting conditions, worn-out coins, and cluttered backgrounds. To overcome these issues, this study introduces an advanced coin recognition system tailored specifically for Indian currency denominations. The research explores several computational approaches, beginning with baseline machine learning models, including Multinomial Naïve Bayes (MNB), Decision Tree Classifier (DTC), and Random Forest Classifier (RFC). Although these models achieved moderate accuracy levels ranging from 70% to 85%, they were unable to effectively capture the subtle variations present in worn and circulated coins. To overcome these challenges, a Deep Convolutional Neural Network (DCNN) architecture was developed. By leveraging hierarchical spatial feature extraction along with Adaptive Feature Enhancement (AFE), the proposed deep learning model achieved a classification accuracy of 100%. The system is further integrated into a dual-access Graphical User Interface (GUI) developed using Tkinter, consisting of an ADMIN module for model training, testing, and evaluation, and a USER module for real-time prediction and result visualization. Model performance was evaluated using key metrics such as precision, recall, and F1-score. The proposed framework offers a scalable, non-intrusive, and efficient solution for banking and retail environments, transforming manual currency handling into a streamlined automated process.

Keywords : Coin Recognition, Deep Learning, Deep Convolutional Neural Network (DCNN), Adaptive Feature Enhancement (AFE), Image Classification.

Author : C. Vijaya Raj, Gadimundati Archana, Ane Charanya, Veluru Chengal Rayulu

Title : Learning-Based Visual Discrimination of Indian Coinage under Illumination and Surface Degradation Variability

Volume/Issue : 2026;03(04)

Page No : 869-878

Building a compelling brand identity is essential for achieving competitive advantage, yet producing unique, meaningful, and market-aligned brand names remains a difficult and time-intensive process. Traditional techniques rely heavily on human creativity, brainstorming practices, and predefined heuristics, which are often subjective, inconsistent, and incapable of capturing deeper semantic relationships or emerging market patterns. This study presents an AI-driven system that integrates natural language processing (NLP) and machine learning to automate and enhance both the generation and evaluation of brand names. The methodology begins with preprocessing textual data through cleaning, tokenization, lemmatization, and removal of stop words to ensure high-quality input. It then utilizes transformer-based embeddings such as DistilRoBERTa to extract rich contextual and semantic representations of words and phrases. For classification tasks, multiple algorithms including Logistic Regression (LR), Random Forest Classifier (RFC), and Support Vector Machines (SVM) are implemented and trained on balanced datasets, where Synthetic Minority Over-sampling Technique (SMOTE) is applied to address class imbalance issues. The system supports both single and batch processing modes, delivering outputs such as predicted brand categories, relevance scores, and performance metrics through an interactive web interface. By combining advanced NLP techniques with robust machine learning models, the proposed platform improves efficiency, scalability, and accuracy while significantly reducing manual effort, bias, and time in developing impactful and market-ready brand names.

Keywords : Brand name generation, Natural Language Processing (NLP), transformer embeddings, DistilRoBERTa, semantic analysis, brand identity

Author : M. Rakesh, Nune Neha, Chitimilla Pavan venkat, Bandaram Pavan kalyan

Title : An AI-Driven Approach to Intelligent Naming System Design and MarketRelevance Evaluation

Volume/Issue : 2026;03(04)

Page No : 879-887

Water resource management (WMT) has become increasingly important due to rising demand, uneven distribution, and concerns regarding water quality and transparency. Many systems rely on centralized storage and manual monitoring, which lead to inefficiencies, lack of accountability, and vulnerability to data tampering, while also lacking strong authentication mechanisms and real-time validation of water quality. These limitations make allocation and billing processes less reliable and expose sensitive data to unauthorized access. To overcome these issues, there is a need for a secure, intelligent, and automated system that ensures data integrity, controlled access, and accurate decisionmaking. The proposed system integrates Blockchain Technology, Machine Learning (ML), and a secure login module within a web-based environment. The authentication mechanism ensures that only authorized administrators can access key functionalities such as water allocation, billing, and payment management, thereby enhance security and prevent unauthorized usage. Blockchain technology is utilized to store all resource and billing records in a decentralized and tamper-resistant manner, ensuring transparency, traceability, and trust across operations. Additionally, a ML based model Random Forest Classifier (RFC) is employed to predict water quality using parameters such as pH, conductivity, turbidity, and organic carbon before allocation decisions are made. This enables proactive assessment of water safety and improves resource utilization. The integration of secure authentication, predictive analytics, and immutable storage enhances system reliability and efficiency. The system supports better decision-making, prevents data manipulation, and ensures accurate and transparent water distribution and billing, making it a robust solution for modern water management applications.

Keywords : Water resource management (WMT), Blockchain Technology, Machine Learning (ML), Smart Water Allocation, Billing and Payment System.

Author : B. Laxmi Pathi, Shambuni Rakesh Goud, Manugonda Rakesh, Boda Rahul

Title : A Trust-Anchored Computational Framework for Real-Time Water Quality Inference with Immutable Consumption Accounting

Volume/Issue : 2026;03(04)

Page No : 888-897

The rapid digital transformation of healthcare, driven by the adoption of Electronic Health Records (EHRs), interconnected medical devices, and integrated information systems, has significantly enhanced operational efficiency and patient care. However, this increased interconnectivity has also expanded the attack surface, exposing healthcare infrastructures to a wide range of cybersecurity threats. Factors such as legacy system dependencies, misconfigurations, and delayed software updates further intensify the risk of data breaches and service disruptions. Traditional manual methods for analyzing security logs and incident reports are increasingly inadequate due to the growing volume and unstructured nature of data, leading to delayed and error-prone threat detection. To address these challenges, this work introduces an automated framework based on Natural Language Processing (NLP) for efficient cyber threat identification and vulnerability analysis. The approach begins with robust text preprocessing, including normalization, tokenization, lemmatization, and removal of irrelevant terms to enhance data quality. Contextual feature extraction is performed using the transformer-based RoBERTa model, enabling deeper semantic interpretation of cybersecurity text. To mitigate class imbalance and improve model robustness, the Adaptive Synthetic Sampling (AdaSYN) technique is applied. The system evaluates multiple machine learning classifiers, including Greedy Tree Classifier (GTC), Tao Tree Classifier (TTC), K-Nearest Neighbors (KNN), and Gaussian Naive Bayes (GNB), facilitating comprehensive performance comparison. Experimental results identify the optimized GTC model as the most effective in classifying threat types, assessing severity, and recommending mitigation strategies. The solution is implemented as a secure web-based application that supports streamlined data input, automated analysis, and intuitive result visualization. The proposed framework enhances detection accuracy, ensures interpretability, and enables proactive cybersecurity management, thereby strengthening the resilience of modern healthcare systems.

Keywords : Electronic health records, Cybersecurity threats, Vulnerability Assessment, Real-Time Threat Analysis, Natural Language Processing (NLP).

Author : M. Rakesh, Gangadhari Vivek, Kashireddy Varun Reddy, Alpana Venkat Ashish

Title : A DeBERT and Greedy Tree Classifier Framework for Multivariate Cyber Threat Prediction in Healthcare Security

Volume/Issue : 2026;03(04)

Page No : 898-910

The rapid growth of electricity consumption and industrial development has significantly increased carbon emissions, raising global concerns about environmental sustainability. Over the years, energy monitoring systems, smart grids, and digital data collection technologies have enabled organizations to gather large volumes of electricity-related data from power plants, distribution networks, and environmental monitoring systems. Analyzing this data has become essential for understanding patterns related to carbon emission reduction and improving energy management strategies. Traditionally, electricity consumption analysis relied on manual methods and basic statistical techniques where analysts used spreadsheets and historical reports to interpret energy usage patterns. However, these traditional approaches were time-consuming and unable to effectively process large and complex datasets generated by modern energy systems. They often failed to capture complex relationships between electricity consumption, environmental conditions, and emission reduction levels, leading to inefficient analysis and limited insights. Therefore, there is a need for advanced analytical techniques capable of handling largescale electricity datasets and accurately identifying emission reduction patterns. In this study, machine learning techniques are applied to analyze electricity consumption data and classify carbon emission reduction categories using models such as Logistic Regression (LR), XGBoost Classifier (XGBC), and Extra Trees Classifier (ETC). In addition, a hybrid analytical approach called the Neuro-Tree Fusion (NTF) model is utilized, which integrates a Graph Polynomial Neural Network (GPNN) for deep feature extraction with a Deep Neural Decision Tree (DNDT) tree-based classifier for final prediction.

Keywords : Electricity Consumption Analysis, Machine Learning, XGBoost Classifier, Neuro-Tree Fusion (NTF), Graph Polynomial Neural Network (GPNN), Deep Neural Decision Tree (DNDT).

Author : J. Naresh, Pendem Bunny, Mohammed Abdul Hashid, S. Jeevan Sagar

Title : A Multi-Model Fusion Approach for Predicting Carbon Emission Behavior in Carbon-Constrained Energy Systems

Volume/Issue : 2026;03(04)

Page No : 911-923