Journal of Science Engineering Technology and Management Science

India is one of the largest producers of fruits and vegetables globally, contributing about 14% of global production., a significant portion of this produce, approximately 30-40%, is lost due to inefficient storage and transportation systems, resulting in an economic loss of ?92,651 crores. The objective is to develop a deep learning-based model that uses temperature simulation data to accurately predict the shelf life of fresh fruits and vegetables, ensuring optimized transport and storage conditions. The title refers to a machine learning-based approach, specifically using deep CNN models, to predict how long fresh produce (fruits and vegetables) will remain viable based on temperature data collected during storage and transport. This system helps in making decisions to reduce spoilage and improve logistics. Traditionally shelf-life predictions relied on fixed temperature guidelines, manual quality checks, and general estimations from historical data, often leading to inaccuracies and higher wastage. Traditional systems lack precision and adaptability, resulting in inefficient management of fresh produce, leading to higher spoilage and economic losses during transportation and storage. This calls for a more dynamic approach. Reducing post-harvest losses is critical to ensuring food security, especially in India. Machine learning models can offer precise shelf-life predictions by incorporating real-time data, thereby reducing waste, improving economic returns, and ensuring fresher produce reaches consumers. A deep learning-based model, using a CNN trained on temperature data collected from storage and transport conditions, can predict the shelf life of fresh produce in real-time. This would allow stakeholders to adjust storage temperatures and optimize distribution routes, ensuring minimal wastage. The AI model can also provide alerts for optimal consumption times and transport decisions. This system will significantly improve efficiency and reduce the cost associated with spoilage

Keywords : Shelf-life prediction, deep learning, CNN, temperature simulation, fresh produce, fruits and vegetables, spoilage reduction, real-time monitoring, storage optimization, transport logistics, postharvest losses, food security, India, AI alerts, economic efficiency

Author : K. Sunil Kumar, Ravinder Reddy.G, Ranjeeth Nadukuda, Sai Kumar Deekonda

Title : DEEP CNN MODEL FOR PREDICTING SHELF LIFE OF FRESH FRUITS AND VEGETABLES USING TEMPERATURE SIMULATION DATA FOR OPTIMIZED TRANSPORT

Volume/Issue : 2025;02(7S)

Page No : 244-252

The global delivery industry faces immense challenges in supply chain inefficiencies, with over 1.3 billion tons of food wasted annually, amounting to 30% of global food production, and resulting in an estimated $940 billion in economic losses each year. Despite these staggering numbers, 40–50% of inventory decisions are still based on manual forecasting techniques, contributing to frequent overstocking or understocking, excessive spoilage, and logistical delays. Traditional forecasting systems are often static, failing to adapt to dynamic market and environmental variables, making them unreliable for real-time driver demand prediction. To overcome these limitations, this research presents a robust regression-based time series forecasting framework aimed at optimizing delivery supply chain management. The proposed methodology utilizes a publicly available driver demand dataset containing temporal sales patterns, weather data, and promotional activity. The dataset undergoes comprehensive preprocessing including normalization, missing value imputation, lag feature creation, and outlier treatment to enhance model input quality. Initially, the Light Gradient Boosting Machine (LGBM) Regressor is employed as the baseline model, demonstrating strong performance. However, to further enhance forecasting accuracy, a Nonlinear Autoregressive model with Exogenous inputs (NARX) Regressor is introduced. NARX leverages both historical internal demand data and external factors (e.g., weather, holidays) for dynamic, multivariate prediction. This model is well-suited for capturing nonlinear dependencies in time series data and supports recursive forecasting for future intervals. Performance evaluation reveals that while the existing LGBM Regressor achieves impressive metrics (MSE: 7.06e-05, MAE: 0.00449, RMSE: 0.00840, R²: 0.948), the proposed NARX Regressor significantly outperforms it with MSE: 6.32e-05, MAE: 0.00350, RMSE: 0.00795, and an R² score exceeding 1.00, indicating superior generalization and near-perfect predictive capability. These results validate the NARX model's effectiveness in minimizing forecasting errors, reducing food waste, and enabling smarter supply chain decisions

Keywords : Dynamic Driver Demand, Ride-Sharing Analysis, NARX Regression, Time Series Forecasting, Supply Chain Management, Delivery Optimization, Food Waste Reduction, LGBM Regressor, Multivariate Prediction, Recursive Forecasting

Author : M. Rakesh, Neha Malyala, Shruthi Peesari, E.Sharanya Reddy

Title : DYNAMIC DRIVER DEMAND IN RIDE-SHARING ANALYSIS USING NARX REGRESSION FOR OPTIMIZED SUPPLY CHAIN MANAGEMENT

Volume/Issue : 2025;02(7S)

Page No : 253-267

Adverse drug reactions pose significant risks in clinical settings, particularly when side effects are overlooked during the early stages of prescription. To mitigate these risks, this study focuses on enhancing drug side effect prediction using machine learning techniques integrated with Explainable AI (XAI) for medical applications. The primary objective is to develop an intelligent, interpretable system that not only predicts potential side effects but also provides transparency in the decisionmaking process, thereby fostering trust among healthcare professionals. A comprehensive dataset comprising drug attributes, side effect profiles, and associated clinical features was used for model training and evaluation. Initial experimentation involved various baseline classifiers, including Ridge Classifier, Linear Support Vector Machine (SVM), Logistic Regression, and Multinomial Naïve Bayes. These models served as benchmarks and were evaluated based on accuracy, precision, recall, and F1-score. Extensive Exploratory Data Analysis (EDA) was performed to uncover patterns, correlations, and class imbalances within the dataset, aiding in informed feature selection and preprocessing. To improve prediction accuracy and enable complex pattern recognition, a Multi-Layer Perceptron (MLP) Classifier was proposed as the advanced model. As a deep learning algorithm, the MLP demonstrated superior performance in capturing nonlinear relationships among features— capabilities that traditional models often lack. This makes the solution suitable for real-world deployment in Clinical Decision Support Systems (CDSS), ensuring safer drug administration and better patient outcomes. The project contributes to the field of medical AI by delivering a highperforming, interpretable, and reliable solution for drug side effect prediction, effectively bridging the gap between complex AI models and practical healthcare applications

Keywords : Adverse drug reactions, Drug side effect prediction, Machine learning, Explainable AI (XAI), Clinical Decision Support Systems (CDSS).

Author : E. Sravanthi, Maddili Vandana, N. Sravanthi, K. Yashashwini

Title : X-DRUGREC: AN EXPLAINABLE AI FRAMEWORK FOR SAFE AND PERSONALIZED DRUG RECOMMENDATION WITH ADVERSE EFFECT PREDICTION

Volume/Issue : 2025;02(7S)

Page No : 268-278

In the era of big data, ensuring data integrity is critical, as even minor corruption can skew analytical outcomes across domains like e-commerce behavior prediction, social network analysis, and intrusion detection systems. Manual data analysis methods often fail to detect subtle yet impactful corruptions, especially in high-dimensional datasets. Additionally, most traditional algorithms lack adaptability to unseen data and often require repetitive retraining. Another limitation is the inability to provide predictive models for future datasets, leaving organizations vulnerable to repeated data integrity issues. Studies indicate that nearly 10–15% of large-scale transactional or behavioral data can be impacted by noise or corruption, while traditional outlier detection methods such as LOF, Isolation Forest, and One-Class SVM achieve only 85–92% accuracy in anomaly detection. Moreover, hybrid supervised approaches like Random Forests, when combined with engineered features, have demonstrated enhanced accuracies exceeding 99% in detecting corrupted data. To overcome these challenges, the proposed PAACDA (Proximity-based Adamic Adar Corruption Detection Algorithm) introduces a novel, graph-theory-based approach using the Adamic-Adar similarity index for anomaly detection. It defines corruption based on local proximity similarity, calculating a PAACDA Index for each data point by measuring deviation from column mean, and setting a range threshold (mean/4). If a data point's similarity is abnormally high, it is flagged as corrupted. Extending this, a Hybrid PAACDA model is proposed by training a Random Forest classifier on features derived from PAACDA, enabling accurate classification of future data instances as normal or corrupted. This hybrid approach not only boosts detection accuracy to 99–100%, but also equips the model with predictive capabilities—an essential step toward proactive data quality assurance in dynamic systems like e-commerce platforms, where customer segmentation and high/low revenue forecasting are vital.

Keywords : PAACDA (Pattern-Aware Adaptive Contextual Data Analysis), Data Corruption Detection, Cyber-Physical Systems, Explainable AI (XAI), Adaptive Modeling

Author : K. Vamshee Krishna, Vishnu Vardhan Reddy, Sunkari Vinay, Yarra Ajay Gopi, K. Gnaneshwar

Title : PAACDA-RF: A HYBRID INTELLIGENCE FRAMEWORK FOR DETECTING DATA CORRUPTION IN COMPLEX CYBER-PHYSICAL SYSTEMS

Volume/Issue : 2025;02(7S)

Page No : 279-288

Botnet attacks, particularly those employing Domain Generation Algorithms (DGA), contribute to over 40% of command-and-control traffic in cyber incidents, with global losses from botnet-driven attacks exceeding $2 billion annually. Traditional detection systems have shown an average false positive rate of 20–30%, significantly impacting response efficiency and operational trust. Studies indicate that explainability in machine learning models improves analyst decision validation by up to 35% in real-world cybersecurity applications. Existing manual detection techniques suffer from limited scalability, high analyst dependency, and lack of consistent interpretability, making them ineffective in real-time or high-volume network environments. Additionally, traditional methods often fail to detect obfuscated or rapidly evolving DGA patterns due to their static rule-based nature and inability to provide insights into the rationale behind decisions. To overcome these challenges, this work proposes an efficient Ensemble Explainable AI (XAI)-based collaborative defense mechanism for botnet detection, with a focus on enhancing interpretability, accuracy, and operational trust. The system utilizes the Botnet DGA dataset and employs a structured pipeline comprising data preprocessing (cleaning, normalization, transformation), SHAP-based feature attribution for transparency, and classification using the XGBoost ensemble model. The model is trained to differentiate between normal and botnet network traffic, with real-time testing supported by the same interpretability mechanisms. The final stage includes a robust performance evaluation using metrics such as accuracy, precision, and recall. Furthermore, collaborative threat intelligence sharing is integrated to refine detection capabilities across organizational boundaries, ensuring adaptive protection against emerging botnet threats while fostering a trusted and explainable cybersecurity ecosystem

Keywords : SHAP (Shapley Additive Explanations), Explainable AI (XAI), DGA (Domain Generation Algorithm), Botnet Detection, Cyber Threat Intelligence

Author : C. Vijayaraj, Vishwith Reddy, L. Manohar Goud, K. Vishnu Vardhan

Title : FEATURE-DRIVEN BOTNET MITIGATION USING EXPLAINABLE SHAP AND XGBOOST ENSEMBLE LEARNING

Volume/Issue : 2025;02(7S)

Page No : 289-310

Modern healthcare systems often rely on centralized storage for Electronic Health Records (EHRs), which introduces several critical vulnerabilities, including single points of failure, potential data tampering, and risks to patient privacy. This research proposes a decentralized EHR management framework that integrates blockchain technology with the Inter Planetary File System (IPFS) to securely handle doctor–patient appointment details, medical reports, prescriptions, and patient feedback. A permissioned Ethereum-like smart contract named “Healthcare” is utilized to immutably store user profiles, appointment metadata, prescription references, and ratings on the blockchain. Sensitive files are encrypted using AES-CTR with keys derived via PBKDF2 and then pinned to a local IPFS node to ensure distributed, confidential file storage. Patients can register, log in, view a list of doctors with on-chain average ratings, book appointments by uploading encrypted reports, and later retrieve prescriptions in decrypted form. Doctors access the system to view appointments, download encrypted medical reports, upload encrypted prescriptions, and publish relevant metadata on-chain. Patient feedback is also recorded as immutable blockchain transactions. An in-memory Python-based caching layer helps improve performance by reducing the need for repeated on-chain reads. Experimental results demonstrate the efficiency of encryption, decryption, and IPFS pinning operations, confirming the system’s performance reliability. This framework strengthens healthcare data management by ensuring transparency, data integrity, and fault tolerance. The combination of blockchain-based metadata storage and off-chain encrypted file hosting through IPFS enables secure, decentralized, and patient-centric healthcare services

Keywords : Electronic Health Records (EHR), Blockchain, IPFS, AES Encryption, PBKDF2, Smart Contracts, Decentralized Storage, Healthcare Privacy, Patient-Centric Systems, Data Integrity, Fault Tolerance, Encrypted File Sharing, Ethereum

Author : Rambabu bommisetti, K. Ushasree, P. Vidhya, G. Vaishnavi

Title : SECURE DECENTRALIZED HEALTH RECORD SYSTEM LEVERAGING BLOCKCHAIN AND IPFS

Volume/Issue : 2025;02(7S)

Page No : 311-318

The exponential growth of Internet-of-Things (IoT) devices has drastically expanded the threat landscape, yet many existing intrusion detection systems overlook the underlying graph structures within device communication networks. This study proposes a comprehensive framework that integrates traditional machine learning models with a graph-based learning strategy enhanced by randomized edge perturbation, aiming to achieve both high detection performance and adversarial privacy protection. The approach begins with the RT_IOT2022 network-flow dataset, performing exploratory data analysis to uncover feature relationships and address class imbalance. Categorical variables such as attack type, protocol, and service are encoded, and SMOTE oversampling is applied to ensure balanced class distribution. Dimensionality reduction using Principal Component Analysis (PCA) yields a compressed yet information-rich feature space. Four baseline classifiers—Gaussian Naive Bayes (GNB), Linear Discriminant Analysis (LDA), Ridge Classifier, and Decision Tree Classifier (DTC)—are trained and evaluated using standard performance metrics. To exploit graph structural information, we propose a hybrid Graph Neural Network–style Decision Tree Classifier (GNN-DTC), where node features are aggregated from local neighborhoods to construct decision trees that capture relational attack behaviors. Additionally, randomized edge perturbation is applied to mask sensitive graph topology and enhance resilience against adversarial attacks, without significantly compromising accuracy. Experimental results reveal that LDA and Ridge achieve accuracies of 93.4% and 90.8% respectively, while GNB trails with 49.0%. The proposed GNN-DTC model achieves 99.7% accuracy and F1-score, substantially outperforming all baselines. Class-wise evaluations further confirm consistent performance across varied attack types. These findings highlight the effectiveness of combining graph-aware learning with privacy-preserving perturbation for secure and robust IoT intrusion detection

Keywords : IoT Security, Intrusion Detection System, Graph Neural Networks, Randomized Edge Perturbation, Adversarial Robustness, Privacy Preservation, Machine Learning, SMOTE, PCA, RT_IOT2022 Dataset, Network Flow Analysis, Graph-Based Learning

Author : K. Vamshee Krishna, Koduri Akarsh, Bethala Harsha, Poosa Vamshi, E Sainath Reddy

Title : SECURE GNNS: DEFENDING GRAPH DATA PRIVACY WITH RANDOMIZED EDGE PERTURBATIONS

Volume/Issue : 2025;02(7S)

Page No : 319-325

Fish disease classification plays a vital role in the success and sustainability of seafood aquaculture, especially in countries like India where the sector makes a substantial economic contribution. Traditionally, disease identification has relied on manual inspection and visual assessment by aquaculture experts. However, these conventional methods are often time-consuming, inconsistent, and heavily reliant on subjective human expertise, leading to potential delays and inaccuracies in diagnosis. To overcome these limitations, the present study focuses on developing automated fish health monitoring systems using cameras and sensors for real-time surveillance. By integrating image processing techniques with deep learning and transfer learning algorithms, the proposed system aims to significantly improve the precision and speed of fish disease classification. The implementation of such smart aquaculture systems marks a transformative step in aquaculture management. These intelligent solutions not only enhance early disease detection and response but also promote more sustainable farming practices by reducing dependency on manual processes. Overall, the use of realtime data analytics and automation contributes to increased productivity, better fish welfare, and lower environmental impact

Keywords : Smart Aquaculture, Fish Disease Detection, Deep Learning, Transfer Learning, Image Processing, Real-Time Monitoring, Automated Aquaculture Systems, Sustainability in Aquaculture, Fish Health Classification

Author : Ritesh Kumar, Gujjula Aravind, A. Ravi Kumar

Title : ENHANCED AQUACULTURE HEALTH MANAGEMENT USING DEEP TRANSFER LEARNING

Volume/Issue : 2025;02(7S)

Page No : 326-332

The history of e-government services dates back to the 1990s when governments began adopting digital technologies to improve public administration and offer services to citizens more efficiently. Initially, egovernment systems were designed to automate administrative tasks, making government services more accessible and reducing the time and cost of manual processing. However, these systems were often limited in their scope, and they struggled with providing personalized services, managing large volumes of data, and integrating various government departments. Despite progress, traditional e-government systems were often criticized for being inflexible, inefficient, and unable to respond rapidly to changing needs. The traditional egovernment approach often relied on static, form-based systems, where citizens had to interact with government websites to complete transactions or submit requests. These systems lacked interactivity, adaptability, and the capacity to handle complex queries or offer tailored services. Additionally, many government agencies still operated in silos, making it difficult for citizens to access comprehensive services from a single platform. Traditional systems were also often plagued by issues related to data security, privacy, and digital divide, leaving certain segments of the population underserved. Artificial Intelligence (AI) presents an opportunity to revolutionize e-government services by addressing these limitations. AI can enhance the efficiency, responsiveness, and accessibility of public services. Through machine learning, natural language processing, and automated decision-making, AI-powered e-government systems can provide personalized, real-time services, improve data analysis, and foster greater citizen engagement. AI can also integrate disparate government systems, enabling seamless service delivery across different departments.The motivation behind this transformation is to provide more efficient, user-centric government services that reduce costs and enhance citizen satisfaction. The proposed AI-based e-government system aims to overcome the limitations of traditional systems, offering more accessible, personalized, and efficient services while improving governance, transparency, and responsiveness to the needs of citizens

Keywords : E-Government Services, Public Service Automation, Smart Governance, AI-Driven Citizen Engagement, Citizen-Centric Services

Author : P. Kamaraja Pandian, Dandi Gunashekar, Prem Kumar, Arun Kumar Yadav

Title : Artificial Intelligence for Streamlined E-Government Services

Volume/Issue : 2025;02(7S)

Page No : 342-350

The classification of animal vocal emotions is a burgeoning field with significant implications for animal welfare, veterinary diagnostics, and conservation, driven by the need to accurately interpret emotional states from vocalizations. Over 500,000 vocal samples across 200 species are now cataloged in repositories like the Animal Vocalization Database (AVD). However, existing manual analysis methods, such as spectrographic analysis and acoustic feature scoring, suffer from subjectivity, with inter-observer agreement averaging only 60%, scalability issues due to time-intensive processes, and noise interference reducing accuracy by up to 30% in field settings. To address these challenges, this study proposes a Deep Learning Convolutional Neural Network (DLCNN) for classifying four emotional classes—anger, disgust, fear, and purr—in animal vocalizations. Preprocessing involves Mel-Frequency Cepstral Coefficient (MFCC) feature extraction, capturing spectral characteristics with 13–40 coefficients per frame, followed by noise reduction and normalization to ensure robustness. Existing methods evaluated include Support Vector Machine (SVM), achieving 82% accuracy; K-Nearest Neighbors (KNN), with 78% accuracy; Decision Tree Classifier (DTC), AdaBoost, and Linear Discriminant Analysis (LDA). These methods struggle with high-dimensional MFCC data and complex emotional patterns, often overfitting or underperforming on noisy inputs. The proposed DLCNN architecture integrates multiple convolutional layers to extract spatial hierarchies, followed by dense layers for classification, trained with a categorical cross-entropy loss function and optimized using Adam. The DLCNN leverages data augmentation (e.g., pitch shifting, time stretching) to enhance generalization, achieving perfect performance metrics: 100% accuracy, precision, recall, and F1-score across all classes. This approach outperforms existing methods by capturing intricate acoustic patterns, offering a scalable, automated solution for real-time animal vocal emotion classification

Keywords : Vocal Emotion Prediction, MFCC Feature Extraction, Animal Sound Classification, Bioacoustic Signal Processing, Spectrogram Analysis

Author : R. Dinesh Kumar, I. Bhavya Sri, G. Shiva Kumar, B. Lavan Kumar, O. Anil Kumar

Title : Prediction of Animal Vocal Emotions using Convolutional Neural Network

Volume/Issue : 2025;02(7S)

Page No : 351-357

ADHD (Attention-Deficit/Hyperactivity Disorder) is a neurodevelopmental disorder affecting children worldwide, characterized by inattention, hyperactivity, and impulsivity. In India, ADHD prevalence ranges between 5% and 15% among school-aged children. Despite increasing awareness, early diagnosis remains a challenge due to social stigma and a lack of standardized screening methods. Traditional diagnosis relies heavily on clinical observation and questionnaires, leading to potential biases and inconsistencies. This study introduces a novel ADHD detection system that integrates a user-friendly graphical user interface (GUI) with advanced machine learning models, particularly emphasizing a Logistic Regression Classifier (LRC) to achieve superior performance. The methodology innovates by combining robust data preprocessing (shuffling and normalization) with a balanced 80:20 train-test split of a 496-record dataset containing 36 movement features extracted from behavioral data. Unlike traditional methods, the system processes both image and video inputs, enabling dynamic and real-world applicable ADHD classification. By leveraging LRC’s ability to model complex relationships in movement data, the system outperforms Naive Bayes (NBC) and Support Vector Machine (SVM), addressing limitations in feature independence assumptions and computational complexity. This approach enhances diagnostic consistency and supports early intervention by providing a scalable, accurate, and practical tool for ADHD detection. The proposed system demonstrates exceptional performance, with the LRC achieving a 97.0% accuracy, 97.0713073% precision, 96.6162253% recall, and 96.8324359% F-score, surpassing NBC (91.0% accuracy) and SVM (94.0% accuracy) on the test set of 100 records. The system’s ability to predict ADHD from image and video inputs via the GUI highlights its practical utility in clinical settings, offering a reliable, automated alternative to subjective assessments

Keywords : ADHD Detection, Pose Estimation, Human Pose Recognition, Behavioral Disorder Diagnosis, Healthcare AI

Author : P. Vyshali, Maddali Datta Sai, P. Shiva Sai Karthikeya, Nidhish Varkala

Title : SmartPose-ADHD: A Supervised Learning Framework for Early ADHD Detection via Pose Estimation

Volume/Issue : 2025;02(7S)

Page No : 358-366

Sentiment analysis is vital for interpreting public opinions on social media platforms like Twitter, where users post in diverse languages that reflect global market trends. Real-time sentiment classification across multiple languages can empower businesses and policymakers with actionable insights, necessitating robust systems capable of handling linguistic diversity and temporal dynamics. The primary challenge lies in accurately classifying sentiments from multilingual tweets in real-time, addressing variations in language (e.g., English, Spanish, French) and cultural nuances in sentiment expression, while ensuring scalability and precision in dynamic global contexts. Conventional sentiment analysis systems typically rely on language-specific, rule-based, or lexicon-based methods that use predefined sentiment dictionaries for a single language, often paired with basic classifiers like Naive Bayes. These traditional approaches struggle with multilingual datasets, requiring separate models for each language, which is inefficient. They also fail to capture contextual and sequential patterns in realtime data, perform poorly on imbalanced datasets, and are unable to handle cross-linguistic sentiment variations effectively—resulting in unreliable predictions for global applications. There is an urgent need for an integrated system capable of real-time sentiment classification across multiple languages by leveraging advanced models to address both contextual and temporal challenges. This project processes multilingual tweets using NLTK for preprocessing (including lowercasing, stopword removal, and stemming), TF-IDF vectorization, and classification models such as Decision Tree, Random Forest, and MLP (despite the title, RNN is not implemented). The system handles a dataset of 17 tweets in English, Spanish, French, German, and Italian, with sentiment labels ranging from 1 to 5 stars, achieving 87.89% accuracy with MLP through a GUI-driven pipeline. The system represents a step forward in multilingual sentiment analysis by offering a user-friendly tool for global sentiment monitoring, though it is currently limited by the small dataset and the absence of RNN. Future enhancements using larger datasets and RNN integration could significantly improve its real-time applicability and robustness for global market in

Keywords : Sentiment Analysis, Real-Time Classification, Natural Language Processing (NLP), Text Preprocessing, Temporal Sequence Modeling

Author : Ch. Mounika, Puppala Mahesh, Pailla Akhil Thomas Reddy

Title : Dynamic Real-Time Sentiment Classifier Using RNNs for Multilingual Social Media Analysis in Global Markets

Volume/Issue : 2025;02(7S)

Page No : 367-374

Noise reduction plays a vital role in enhancing audio clarity across telecommunications and broadcasting systems, where maintaining speech intelligibility under diverse noise conditions is essential. Traditional approaches like spectral subtraction rely on estimating and removing stationary noise components from the signal spectrum; while effective in stable environments, they often introduce audible artifacts and perform poorly in dynamic or real-time scenarios involving non-stationary noise. To address these challenges, this work proposes a novel Canonical Correlation Deep Learning (CCDL) method for advanced noise reduction that dynamically adapts to varying acoustic environments. The CCDL approach leverages deep learning capabilities alongside canonical correlation analysis to extract meaningful relationships between noisy and clean signals, enabling intelligent parameter adjustment in real time. Its Canonical Correlation Bandpass (CCB) configuration focuses on enhancing critical speech frequency bands while suppressing irrelevant noise, thereby preserving the naturalness and integrity of the audio. Designed for real-time adaptability, CCDL minimizes latency and artifacts, making it highly suitable for applications such as mobile communications, live broadcasts, and voice-based interfaces. Comprehensive evaluations using performance metrics such as Signal-to-Noise Ratio (SNR), Perceptual Evaluation of Speech Quality (PESQ), and LogSpectral Distance (LSD) demonstrate the proposed method’s superiority over conventional techniques, highlighting its effectiveness in delivering clearer, more intelligible audio. CCDL marks a significant advancement in adaptive noise reduction, offering a robust, real-time solution to overcome the limitations of traditional systems in modern communication networks.

Keywords : Noise Reduction, Audio Enhancement, Real-Time Adaptability, Signal-to-Noise Ratio (SNR), Log-Spectral Distance (LSD)

Author : M. Amareswar, Nishitha Digadari, Bhagwan Sahu, Aakiti Manikanta Reddy

Title : Advanced Noise Reduction Using Canonical Correlation Deep Learning (CCDL) for Real-Time Audio Enhancement

Volume/Issue : 2025;02(7S)

Page No : 375-381

Life insurance products have evolved over time from being traditional protection-oriented schemes to investment-linked instruments. This study explores the preferences of customers between the two major components of life insurance—protection and investment. The research employs both primary and secondary data to analyze how demographic factors, financial literacy, risk appetite, and life goals influence product selection. A structured questionnaire was administered to 300 policyholders across urban and semiurban areas in Chhattisgarh, India. The results reveal that younger policyholders and those with higher incomes tend to prefer investment-linked policies such as ULIPs, while older or risk-averse individuals opt for term insurance focused purely on protection. Secondary data from IRDAI and LIC reports were also examined to understand broader trends. The findings can help insurance companies design more targeted products, and also assist regulators in encouraging better financial planning among consumers. The study concludes by recommending personalized policy bundling and greater financial awareness to bridge the gap between perceived and actual insurance needs.

Keywords : Life Insurance, Customer Preference, Investment, Protection, ULIPs, Term Insurance, Policyholders, Financial Planning

Author : Dr Rajeshvar Sao

Title : INVESTMENT VS PROTECTION: A STUDY OF CUSTOMER PREFERENCE IN LIFE INSURANCE PRODUCTS

Volume/Issue : 2025;02(7S)

Page No : 382-387

The Capital market is a market for financial assets which have a long or indefinite maturity. Generally, it deals with long term securities which have a maturity period of above one year. Capital market may be further divided into three types Industrial securities market, Government securities market and Long term loans market. Industrial securities market is further divided into two types primary market or new issue market and secondary market or stock exchange. Government securities market is also called as Gilt-Edged securities market. It is the market where Government securities are traded. Long term loans market is divided into three types Term loans market, Mortgages market and Financial guarantees market. Absence of capital market instruments acts as a deterrent to capital formation and economic growth. Resources would remain idle if finances are not funneled through the capital market. a) The capital market instruments serves as an important source for the productive use of the economy?s savings. It mobilizes the savings of the peoples for further investment and thus avoids their wastage in unproductive uses. b) It provides incentives to saving and facilitates capital formation by offering suitable rates of interest as the price of capital. c) It provides an avenue for investors, particularly the household sector to invest in financial assets which are more productive than physical assets. d) It facilitates increase in production and productivity in the economy and thus, enhances the economic welfare of the society. Thus, it facilitates “the movement of stream of command over capital to the point of highest yield” . e) A healthy capital market instrument consisting of expert intermediaries promotes stability in values of securities representing capital funds.

Keywords :

Author : Kadila likitha , Dr Mohd Abdul Hafeez

Title : A STUDY ON CAPITAL MARKETS AT BAJAJ CAPITAL LTD

Volume/Issue : 2025;02(7S)

Page No : 388-395

Most portfolio analysis is based on the use of two parameters, the mean and variance, of the statistical distribution of returns. Exceptions to this practice can be found in an empirical work both using the third moment around the mean. It is the purpose of this project is to begin a general extension of the two-parameter analysis to three or more parameters. Accordingly, some problems will be solved, but others will be suggested for further analysis. Portfolio analysis helps you decide which of these products and services should be emphasized and which should be phased out, based on objective criteria. Portfolio analysis consists of subjecting each of the association's products and services through a progression of finer screens. During a time of cutbacks and scarce resources, it is essential to screen out programs and services that are not essential to most members. Those that appeal to a more limited segment can be funded by those desiring the product or service rather than by dues.

Keywords :

Author : Kanaka Manognya, Dr K MADHU BABU

Title : PORTFOLIO MANAGEMENT AT KARVY STOCK BROKING LTD

Volume/Issue : 2025;02(7S)

Page No : 396-405

In the usual course of business, assets are not retained with the idea of resale; rather, they are held with the intention of being used continuously to produce or provide goods or services. Things like land, buildings, machinery, cars, furniture, and fixtures are examples. To accurately assess the company's financial health and profitability, it is necessary to value its fixed assets. Many years of service are anticipated from fixed assets. Whether due to use, the passage of time, or some other factor, the value of these assets declines over time. Depreciation allows for the reduction of fixed assets when they are used up and their value is turned into cash. Since depreciation is a component of overall production costs, accurate income calculation requires accurate depreciation measurement. The records are kept in order with the going concern element of the business unit in mind when financial transactions are documented. According to the going concern analysis, a company can keep making money as long as it has a realistic expectation of being in business forever. Much of the rationale for methodically depreciating fixed assets from their original cost rather than their present realizable value is based on this premise.

Keywords :

Author : JYOTHIKA , Dr G Sudhakar

Title : FIXED ASSET MANAGEMENT AT BAJAJ ELECTRICALS LTD KANNEBOINA

Volume/Issue : 2025;02(7S)

Page No : 406-411

This study examines the recruitment and selection practices at Karvy Stock Broking Ltd, a leading financial services company. Effective recruitment and selection processes are vital for organizations to attract, retain, and develop a talented workforce. The research explores the strategies employed by Karvy Stock Broking Ltd to identify, attract, and select suitable candidates for various roles within the organization. The study utilizes a mixed-methods approach, combining qualitative interviews with HR personnel and employees, as well as quantitative analysis of recruitment metrics and performance data. Key aspects investigated include the alignment of recruitment practices with organizational goals, the use of technology in the recruitment process, the assessment of candidate competencies, and the evaluation of the effectiveness of selection methods. Findings reveal that Karvy Stock Broking Ltd emphasizes the importance of aligning recruitment practices with the company's strategic objectives. The organization utilizes a combination of traditional methods and innovative technologies, such as online job portals and social media platforms, to attract a diverse pool of candidates. Additionally, the study highlights the significance of competency-based assessments and structured interviews in identifying candidates who fit the company culture and possess the required skills and attributes. The research provides valuable insights into the recruitment and selection processes at Karvy Stock Broking Ltd and offers recommendations for enhancing these practices to ensure the continued success of the organization in attracting and retaining top talent.

Keywords :

Author : POOJITHA, (Dr) R Srinivasa Rao

Title : RECRUITMENT AND SELECTION AT KARVY STOCK BROKING LTD KANUR

Volume/Issue : 2025;02(7S)

Page No : 412-414

In day-to-day management of the firm, it is essential to manage the inventory so as to maintain proper supply of goods at proper time. Inventory represents an important decision variable at all stages of product manufacturing, distribution and sales, in addition to being a major portion of current assets of many organizations. Too much and too low inventories bring down the level of profitability of an organization. Therefore, whether it is a manufacturing or merchandized organization, the goal should always be the same that is, to ensure the inventory is ready and at the same time inventory is at a low level. Inventory management is functional field of finance and production that covers the efficient and effective use of raw materials and spares which are consumed in producing the finished goods in manufacturing concern. A firm ignoring the management of inventories will be jeopardizing its long run profitability and may fail finally. The reduction in „excessive? inventories carries a favourable impact on a company?s profitability. This paper consists of different parts where the inventory management concepts are discussed, and different inventory control techniques are discussed. This paper also introduces the various costs incurred due to the storage inventory, economic order quantities, stock levels, shortage costs, inventory methods.

Keywords :

Author : Kasagoni swetha ,Dr Chandra Sekhar Pattnaik

Title : BUDGETARY CONTROL AT COCA COLA

Volume/Issue : 2025;02(7S)

Page No : 415-423

To many among the general public, the term volatility is simply synonymous with risk associating with any financial asset. Volatility plays a vital role in the secondary market which influences a lot on the investment decision. The main objective of this study is to examine the aforesaid two aspects of the volatility to curb the excess volatility in the market. In this paper an earnest attempt is made to know the tendency of inter and intra- day-volatility in Indian

Keywords :

Author : Katha Sneha , Dr Mohd Abdul Hafeez

Title : VOLIATILITY OF SELECTED BSE STOCKS AT INDIA INFOLINE LTD

Volume/Issue : 2025;02(7S)

Page No : 424-433

Taking into account the rapidly changing of the environment nowadays and the necessity of being different between organizations, this project tries to show how to achieve a sustainable competitive advantage in companies, through talented people using talent management strategies. Here is included all theoretical framework where we will explain our understanding of talent management, talented people and the creativity as a talent. This framework gives us the tools needed to be able to analyse a real talent management strategy. During the analysis we will discover that a talent management strategy has to be fitted with the corporate strategy and with the corporate culture and also, that there are infinite ways to develop the talent management activities, it depends on the organization which develops it. For instance we are going to study Varun Motors which are very different between them but both of them have a talent management strategy.

Keywords :

Author : Katkuri Nikhila, Dr D Sushma Kiran

Title : TALENT MANAGEMENT AT KOTAK MAHINDRA BANK

Volume/Issue : 2025;02(7S)

Page No : 434-441

The study deals with Employee’s Perception towards Corporate Social Responsibility. The research methodology of primary and secondary data has been adopted along with objectives of the study. Data were collected through structured questionnaire from 100 Respondents. An elaborate discussion on concept of CSR is discussed. The framework was made to proceed with the work and to bring the study to a conclusion. This study summarizes the findings emerged from analysis of the results pertaining to various aspects on the corporate social responsibility in organizations. Few suggestions have been given for organizations to improve the awareness level of employees on CSR

Keywords :

Author : Kavampally renuka, Dr M Narasimha

Title : EMPLOYEE AWARENESS TOWARDS CORPORATE SOCIAL RESPONSIBILITY AT RELIANCE RETAIL LTD

Volume/Issue : 2025;02(7S)

Page No : 444-449

A capital expenditure is an outlay of cash for a project that is expected to produce a cash inflow over a period of time exceeding one year. Examples of projects include investments in property, plant, and equipment, research and development projects, large advertising campaigns, or any other project that requires a capital expenditure and generates a future cash flow. Because capital expenditures can be very large and have a significant impact on the financial performance of the firm, great importance is placed on project selection. This process is called capital budgeting. Potentially, there is a wide array of criteria for selecting projects. Some shareholders may want the firm to select projects that will show immediate surges in cash inflow, others may want to emphasize long-term growth with little importance on short-term performance. Viewed in this way, it would be quite difficult to satisfy the differing interests of all the shareholders. Fortunately, there is a solution. The goal of the firm is to maximize present shareholder value. This goal implies that projects should be undertaken that result in a positive net present value, that is, the present value of the expected cash inflow less the present value of the required capital expenditures. Using net present value (NPV) as a measure, capital budgeting involves selecting those projects that increase the value of the firm because they have a positive NPV. The timing and growth rate of the incoming cash flow is important only to the extent of its impact on NPV. Using NPV as the criterion by which to select projects assumes efficient capital markets so that the firm has access to whatever capital is needed to pursue the positive NPV projects. In situations where this is not the case, there may be capital rationing and the capital budgeting process becomes more complex. This idea is crucial in the principal-agent relationship that exists between shareholders and corporate managers. Even though each may have their own individual preferences, the common goal is that of maximizing the present value of the corporation.

Keywords :

Author : Ketham Soniya , Dr K MADHU BABU

Title : CAPITAL MARKET AT KESORAM CEMENT LTD

Volume/Issue : 2025;02(7S)

Page No : 450-455

Capital market has emerged as a separate academic discipline in India. Portfolio theory that deals with the rational investment decision-making process has now become an integral part of financial literature. This study evaluates the investment potential of five companies in the context of their average returns, volatility, systematic risk, and expected returns. The companies analyzed include Tata Steel, Bajaj Auto, Bharat Dynamics, TVS Electronics, and Vaibhav Global. Findings reveal negative average returns across all companies, coupled with varying degrees of volatility and systematic risk. Recommendations suggest exercising caution and monitoring closely for signs of stability before considering investment. The conclusion emphasizes the importance of prudent observation and patience, highlighting the discrepancy between expected and actual returns. Ultimately, waiting for improved performance could mitigate risks and lead to more informed investment decisions.

Keywords :

Author : Kodepaka prathyusha ,Dr G Sudhakar

Title : CAPITAL MARKET AT ICICI DIRECT LTD

Volume/Issue : 2025;02(7S)

Page No : 451-457

There are multile technical analysis tools available for security analysis such as MACD,RSI, MFI,ADX and candle stick pattern etc. but still investors or traders are confused about accuracy of indicators and pattern. This research focuses to test accuracy of signals given by MFI indicator and also compared to the percentage of correctness towards buy and sell signals. This research considered secondary data of twelve banks during the period of Apr 2022 to June 2022 in which all the selected banks are from “bank nifty index”. In the study it has been measured that the signals given by MFI indicator were not significant and there is significant difference in correctness of Buy signals and Sell signals as per MFI indicator. It is also observed that MFI indicator provide more sell signals as compare to buy signals. Hence, it can be concluded that the use of only MFI indicator will not be appropriate for the investor for decision making but this indicator would be more useful for selling decision rather than buying decision.

Keywords :

Author : Kodirekka Esther , Dr G Sudhakar

Title : CASH FLOW MANAGEMENT AT ICICI SECURITIES LTD

Volume/Issue : 2025;02(7S)

Page No : 458-465