Journal of Science Engineering Technology and Management Science

Rapid urban transformation across the Balkan region has intensified the interaction between cities, capital inflows, and growth trajectories, particularly through real estate–led development strategies. The purpose of this study is to examine how foreign direct investment (FDI) influences urban growth patterns by channeling capital into real estate markets and reshaping the spatial and economic structure of Balkan cities. The study adopts a quantitative research design using panel data drawn from selected Balkan economies over the period 2005–2023. Secondary data sourced from international databases and national urban development reports are analyzed using panel regression and mediation techniques to assess the dynamic relationship between FDI inflows, real estate investment, and urban economic growth. The findings reveal that FDI exerts a statistically significant and positive effect on urban growth, with real estate investment functioning as a critical transmission mechanism. Cities attracting sustained capital inflows experience accelerated infrastructure development, rising property values, and expanded construction activity, which collectively contribute to short- and medium-term economic growth. However, the results also indicate uneven development outcomes, highlighting emerging risks related to speculative investment and spatial inequality. The originality of this study lies in its city-level empirical focus on the Balkans, a region that remains underexplored in urban economics and investment literature. By integrating capital dynamics with real estate–driven urbanization, the study offers a connected analytical framework linking global investment flows with local urban transformation. The results provide valuable insights for policymakers, urban planners, and investors seeking to balance growth, sustainability, and inclusiveness in rapidly transforming Balkan cities.

Keywords : Channelling, Speculative, Estate–Driven, Estate–Driven, FDI, Real estate, Urban development, Urban Economics, Investment

Author : Dr. MUHAMMAD ALI SHAHZAD

Title : A RESEARCH PAPER ON CITIES, CAPITAL AND GROWTH: REAL ESTATE–LED URBAN DEVELOPMENT AND FDI DYNAMICS IN THE BALKANS

Volume/Issue : 2026;03(03)

Page No : 1-15

This study focuses on the evaluation of merchant acquisition & QR Payments Adoption of India post payments bank (IPPB). IPPB plays a significant role in expanding digital financial services by utilizing the vast India post network to reach merchants across rural and semi urban areas. The primary objective of this study is to analyse the extent of QR payment adoption among IPPB merchants and asses the effectiveness of IPPBS’S merchant acquisition strategy. This study examines the role of India post staff and postal network in on boarding merchants and promoting QR- based digital payments. Doorstep banking services, awareness programs, and continuous support provided by postal employees are analysed to understand their impact on merchant participation. The research also evaluates the level of usage, convenience, and benefits experienced by merchants after adopting IPPB QR payments, including ease of transactions, reduced cash dependency, improved transparency, and customer satisfaction. Further, the study identifies various challenges faced by merchants while using IPPB QR payments. These challenges include limited digital literacy, lack of technical knowledge, network connectivity issues, concerns regarding security, and resistance to adopting digital modes of payments. Understanding these issues helps in assessing the gaps in the existing merchant acquisition and digital payments frame work. The study is based on both primary and secondary data. Primary data is collected through structured questionnaires and direct interactions with IPPB merchants, while secondary data is gathered from journals, reports, official IPPB publications, and digital payment studies. The findings reveal that while IPPB has successfully expanded QR payment adoption among merchants, continuous training, awareness initiatives, and technological improvements are necessary for sustained to improve merchant acquisition and enhance the effectiveness of QR payment usage in IPPB, contributing to financial inclusion and a cashless economy nationwide.

Keywords : cashless economy, Indian postal payments bank, merchant acquisition, QR payments, digital literacy, digital payments, rural banking.

Author : C. Hari Chandana, S. Izaz Ahammed

Title : A PROJECT ON MERCHANT ACQUISITION & QR PAYMENTS IN IPPB

Volume/Issue : 2026;03(03)

Page No : 16-28

The rapid growth of electric vehicles (EVs) has significantly increased the demand for fast and efficient charging infrastructure. Multi-port fast EV charging stations enable simultaneous charging of multiple vehicles, but they introduce major challenges such as power imbalance, grid stress, peak load demand, and inefficient energy utilization. To address these issues, this paper proposes an Adaptive Smart Energy Management System (ASEMS) for multi-port fast EV charging stations.The proposed system dynamically allocates available power among multiple charging ports based on real-time parameters such as battery state-of-charge (SoC), vehicle priority, grid availability, and peak demand conditions. An intelligent control algorithm continuously monitors load variations and optimizes power distribution to minimize energy losses, reduce charging time, and prevent grid overloading. The system also integrates renewable energy sources and energy storage systems to enhance sustainability and reliability.Simulation results demonstrate that the proposed adaptive strategy improves energy efficiency, reduces peak demand stress on the grid, and ensures balanced load sharing compared to conventional fixed power allocation methods. The developed framework provides a scalable and cost-effective solution for next-generation multi-port fast EV charging infrastructure.

Keywords : Adaptive Energy Management, Multi-Port EV Charger, Fast Charging Stations, Smart Power Allocation, Load Balancing, Electric Vehicle (EV), DC Fast Charging, Real-Time Energy Optimization, Grid Integration

Author : B. VINOD KUMAR, K. VENKAT RAO, S. SWAPNA, T. ARUN KUMAR

Title : ADAPTIVE SMART ENERGY MANAGEMENT SYSTEM FOR MULTI-PORT FAST EV CHARGING STATIONS

Volume/Issue : 2026;03(03)

Page No : 29-35

Power quality disturbances such as voltage sag, swell, and interruptions significantly affect sensitive loads in power distribution systems. This paper proposes a Dynamic Voltage Restorer (DVR) capable of mitigating these disturbances without employing conventional controllers such as P, PI, PID, fuzzy logic, or neural networks. The proposed DVR consists of a battery energy storage system, Voltage Source Inverter (VSI), LC filter, series injection transformer, and control circuitry for pulse generation. The three-phase supply voltages are transformed into the Direct–Quadrature–Zero (DQ0) reference frame and compared with reference values to generate error signals. Under normal operating conditions, no compensating voltage is injected. During disturbances, the detected error signals are processed through inverse DQ0 transformation to generate switching pulses for the VSI, enabling appropriate voltage injection to restore the load voltage to its rated value. Simulation results obtained in MATLAB/Simulink demonstrate the effectiveness of the proposed DVR in mitigating balanced and unbalanced sags, swells, and single- and three-phase outages.

Keywords : Dynamic voltage restorer (DVR), sag, swells, short interruption, power quality, VSI

Author : G.JAYA CHANDRA, MR.V.V.NARAYANA REDDY, D.YEDUKONDALU,J.V.LOKESH

Title : ENHANCED POWER QUALITY IMPROVEMENT USING ENERGY STORAGE INTEGRATED DYNAMIC VOLTAGE RESTORER

Volume/Issue : 2026;03(03)

Page No : 36-42

The global commitment to mitigating climate change has accelerated the transition toward cleaner and more sustainable energy systems. In this context, green finance has emerged as an important financial mechanism that supports investments in environmentally sustainable projects, particularly in the renewable energy sector. The availability of green financial instruments such as green bonds, sustainable loans, and climate funds has created new opportunities for corporations to allocate capital toward renewable energy initiatives. Despite the growing attention given to sustainable finance, there remains limited empirical evidence on how green finance influences corporate investment decisions in renewable energy projects. This study examines the relationship between green finance mechanisms and corporate investment behaviour in the renewable energy sector. The research focuses on how financial accessibility, policy incentives, and sustainability commitments influence corporate decisions to invest in renewable energy infrastructure. Using empirical data and statistical analysis, the study evaluates the extent to which green financial instruments encourage firms to adopt sustainable investment strategies. The findings are expected to provide insights for policymakers, financial institutions, and corporate managers seeking to strengthen the role of finance in supporting the global energy transition. By highlighting the importance of green finance in facilitating renewable energy investments, the study contributes to the broader discussion on sustainable economic development and corporate environmental responsibility.

Keywords : Green Finance, Energy Transition, Renewable Energy Investment, Corporate Sustainability, Sustainable Finance, Climate Finance.

Author : R. Nagaraju, G. Shivaranjani, E. Chandra Shekhar

Title : DRIVING SUSTAINABLE ENERGY TRANSFORMATION: THE ROLE OF GREEN FINANCE IN CORPORATE RENEWABLE ENERGY INVESTMENTS

Volume/Issue : 2026;03(03)

Page No : 43-56

57-73Since early detection may strongly support ap- propriate treatment decisions, brain tumor analysis utilizing magnetic resonance imaging (MRI) is a crucial duty in medical diagnosis. The demand for automated solutions arises due to the time-consuming nature of manual MRI scan evaluation and its heavy reliance on expert interpretation. This study proposes a computeraided system for brain tumor analysis that integrates a deep learning-based classification method with fuzzy thresholding-based picture segmentation. MRI brain pictures are processed initially to improve clarity and minimize undesired noise. Fuzzy thresholding, which aids in controlling intensity changes found in medical images, is then used to segregate tumor areas. Following segmentation, the pictures are classified into appropriate tumor classifications using a deep learning model. When compared to conventional methods, the experimental results show that the suggested structure offers dependable and consistent performance. Overall, the established method provides an effective and automated way to help medical practitioners identify and categorize brain tumors

Keywords : Terms—Brain tumor detection, MRI analysis, fuzzy thresholding,tumor classification

Author : T.S.R.Krishna Prasad, Uddanti Deepika, Premsagar Pitla,Thimmasarthi Tapaswi,Shaik Imran

Title : Improved Brain Tumor Detection and Classification in Brain MRI using Deep Learning Techniques

Volume/Issue : 2026;03(03)

Page No : 57-73

Investment decision-making in volatile markets presents significant challenges due to the limitations of traditional capital budgeting methods, such as Net Present Value (NPV) and Internal Rate of Return (IRR). These conventional approaches often fail to account for managerial flexibility and strategic adaptability, leading to undervaluation of opportunities under high uncertainty. Real Options Theory (ROT) offers a dynamic alternative by treating investments as options that can be adapted based on market conditions. However, classical real options models typically rely on risk-neutral assumptions, neglecting investor risk preferences and behavioural factors, especially during periods of financial instability. To address these shortcomings, this study develops and empirically examines a risk-adjusted real options valuation (RAROV) framework. By integrating investor-specific risk preferences, subjective discount rates, and stochastic volatility, the proposed model enhances valuation accuracy and strategic decision-making in uncertain environments. Empirical results demonstrate that risk-adjusted frameworks outperform traditional models in terms of valuation precision, investment timing, and risk-adjusted returns. Additionally, advances in computational finance and machine learning further strengthen the applicability of these frameworks, enabling more accurate forecasting and adaptive investment strategies. Ultimately, this research provides a robust, empirically validated approach to capital budgeting, supporting managers and policymakers in crafting resilient and value-enhancing investment strategies under market volatility.

Keywords : Capital budgeting, empirical validation, investment decision-making, stochastic volatility, strategic flexibility

Author : Savanam Chandra Sekhar

Title : RISK-ADJUSTED REAL OPTIONS VALUATION FOR STRATEGIC DECISIONMAKING IN VOLATILE MARKETS

Volume/Issue : 2026;03(03)

Page No : 74-84

Brain tumors are among the most critical and life-threatening medical conditions, making early and accurate detection essential for effective treatment planning. The main objective of this project is to study the application of transfer learning using advanced deep learning architectures such as VGG-16, VGG-19, ResNet-50, Inception-V3, and DenseNet-201 for accurate and efficient brain tumor detection from MRI images. These models help address challenges like limited medical imaging data and high computational requirements. Traditional manual analysis of brain MRI scans is time-consuming and often affected by human subjectivity, which makes automated detection systems highly valuable. The proposed approaches utilize pre-trained deep learning models that are fine-tuned to classify brain tumor images effectively. Each model is evaluated using a benchmark dataset, with preprocessing techniques such as normalization, augmentation, and segmentation applied to improve feature extraction. Performance metrics including accuracy, precision, recall, and F1-score are used to assess and compare the models. Experimental results show that ResNet-50 provides the best performance due to its deeper architecture and strong feature extraction capability, followed by VGG-19 and Inception-V3. DenseNet-201 demonstrates a good balance between computational efficiency and accuracy, while VGG-16 performs reliably despite its relatively simple structure. This research highlights the effectiveness of transfer learning in overcoming challenges related to limited data and computational constraints in medical imaging. The findings provide valuable insights for deploying deep learning models in clinical environments, helping improve diagnostic accuracy and efficiency in brain tumor detection.

Keywords : Brain Tumor; Transfer Learning; CNN; Accuracy; F1 Score

Author : MR.K.L.SUNDEEP,B. PRIYANKA,B. RATNAKAR,K. KEERTHANA,B. RAHUL

Title : BRAIN-TUMOR DISEASE DETECTION

Volume/Issue : 2026;03(03)

Page No : 85-98

The increasing volume and speed of network traffic, along with the growing sophistication of cyber threats, have made real-time intrusion detection essential for modern digital infrastructures. Traditional security systems often depend on static rules or offline analysis, which limits their ability to detect new or rapidly evolving cyberattacks. This paper proposes a hybrid architecture that combines stream processing frameworks with deep learning models to enable real-time threat detection from network logs. The system uses Apache Kafka for efficient log ingestion and Apache Flink for real-time stream processing and analytics. Deep learning models, including Long Short-Term Memory (LSTM) networks and one-dimensional Convolutional Neural Networks (1D CNN), are applied for anomaly detection and threat identification. The proposed approach is evaluated using benchmark datasets such as CIC-IDS 2017 and UNSW-NB15. Experimental results show that the system can detect network threats with high accuracy and low latency while maintaining scalability under highthroughput conditions. The architecture is therefore suitable for deployment in real-world operational environments where fast and accurate threat detection is crucial.

Keywords : real-time threat detection; network log analysis; stream processing; deep learning; anomaly detection

Author : DR.MAHESH,GPVS. KARTHIK,G. KISHORE,B. ABHINAV BALAJI, SUMITH RATHOD

Title : Real-Time Threat Detection Using Stream Analytics and Deep Learning

Volume/Issue : 2026;03(03)

Page No : 99-103

The electrification of transportation systems is increasing rapidly, and city buses offer significant potential in this transition. A deep understanding of real-world driving data is essential for improving vehicle design and optimizing fleet operations. Several technological factors must be considered to operate alternative powertrain systems efficiently. Uncertainty in energy demand often leads to conservative vehicle design, which can result in inefficiencies and higher operational costs. Due to the complexity and interdependence of various parameters, both industry and academic researchers have faced challenges in developing analytical solutions for this problem. Accurate prediction of energy demand can enable cost reduction and improved operational planning. This paper focuses on increasing transparency in the energy consumption behavior of battery electric buses (BEB). A new set of explanatory variables is introduced to describe speed profiles, which are then applied within machine learning models for prediction. The study develops and evaluates five different algorithms in terms of prediction accuracy, robustness, and practical applicability. The proposed models achieved prediction accuracy of more than 94%, showing strong performance when combined with carefully selected features. The presented methodology offers valuable potential for manufacturers, fleet operators, and urban communities to support the transition toward efficient and sustainable public transportation systems.

Keywords : Battery Electric Buses, Energy demand prediction, Feature extraction, Machine learning, Meta modeling

Author : MR. B.RAJU, D. MUKESH TENDULKAR, B.KESHAV, K. GANESH, D. SHIVA SAI

Title : DATA DRIVEN ENERGY ECONOMY PREDICTION FOR ELECTRIC CITY BUSES

Volume/Issue : 2026;03(03)

Page No : 104-115

Organ matching and transplantation are life-saving medical procedures that require accurate donor– recipient compatibility assessment to reduce rejection risks and improve transplant success rates. However, the current organ transplant system faces several challenges, including inefficiencies in donor-recipient matching, concerns about data security, organ trafficking, and limited transparency in allocation procedures. Blockchain technology provides a decentralized, secure, and transparent solution to address these problems. By using blockchain, organ donation registries can be securely maintained, ensuring that donor and recipient records remain tamper-proof and reliable. Smart contracts can automate the matching process based on predefined compatibility criteria, improving efficiency and reducing waiting times for patients. In addition, blockchain technology improves traceability, helping to prevent fraud and unethical activities in organ allocation. This paper explores the integration of blockchain technology in organ matching and transplantation systems, highlighting its potential to enhance data integrity, increase transparency, and streamline the transplant process, ultimately improving trust among stakeholders and saving more lives.

Keywords : Decentralized System, Python (flask), Block chain Technology

Author : MRS.M. KAVITHA, AMRUTHA VARSHINI, Y. CHENNAKESHAVULU, M. SAI SRI, M. PRANEETH

Title : ORGAN MATCHING & TRANSPLANTATION BLOCK CHAIN

Volume/Issue : 2026;03(03)

Page No : 116-122

The increasing frequency and complexity of web attacks require strong security mechanisms to protect modern digital infrastructures. Traditional web attack detection systems mainly rely on predefined rules or signaturebased methods, which can often be bypassed by advanced and evolving malicious techniques. This paper proposes a deep learning-based approach for detecting web attacks using an end-to-end learning framework that improves the identification and prevention of web-based threats. The proposed system utilizes a Deep Neural Network (DNN) to analyze patterns and anomalies in web traffic data. Through end-to-end learning, the model automatically extracts meaningful features from raw input data, eliminating the need for manual feature engineering. This capability allows the system to adapt more effectively to new and previously unseen attack patterns. The model is designed to detect various types of web attacks, including SQL injection, cross-site scripting (XSS), and distributed denial-of-service (DDoS) attacks. The study also discusses important stages such as web traffic data collection, preprocessing, model training, and system optimization. Additionally, the detection model can be integrated with existing web security frameworks to enhance protection mechanisms. By utilizing advanced deep learning architectures such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), the system achieves high detection accuracy and supports real-time threat identification. This research highlights the potential of deep learning techniques in cybersecurity by providing an adaptive and proactive solution capable of evolving with emerging web attack strategies.

Keywords : Web attack detection, Deep learning, Anomaly detection,Cybersecurity,Adaptive defense

Author : MR.V. SRINIVAS, S. ABHINAY, V. BHARGAV, U. SRINIVAS SAGAR, V. SIDDU, R. HARSHAVARDHAN

Title : DETECTING WEB ATTACKS WITH END-TO-END DEEP LEARNING

Volume/Issue : 2026;03(03)

Page No : 123-128

Recent advancements in predictive technologies are improving the understanding and management of Parkinson’s disease, a progressive neurological disorder that affects millions of people worldwide. This study presents a machine learning based approach to predict the progression and severity of Parkinson’s disease using the UCI Parkinson’s Telemonitoring Vocal Dataset. The proposed methodology applies deep neural networks along with machine learning techniques to analyze vocal characteristics associated with the disease. Advanced algorithms such as the Random Forest Classifier are utilized to enhance prediction accuracy and improve disease detection. By analyzing patterns in the dataset, the developed models can assist in forecasting disease severity and identifying symptoms at earlier stages. This approach aims to support healthcare professionals in making more informed diagnostic decisions and planning effective treatment strategies. Ultimately, the proposed system contributes to improving the quality of life for patients with Parkinson’s disease by enabling earlier detection and more precise medical intervention.

Keywords :

Author : MR. DR. AMIT, M. INDHU, V. SAI GANESH, S. ADITHYA, MONISH RAJU

Title : PREDICTION OFPARKINSON'S DISEASE AND SEVERITY OF THE DISEASE USING ML AND DL ALGORITHMS

Volume/Issue : 2026;03(03)

Page No : 129-135

The proposed research project presents an integrated machine learning framework for detecting fraudulent activities on online job platforms using Random Forest and XGBoost algorithms. The system applies a comprehensive multi-layered approach that combines natural language processing, feature engineering, and statistical analysis to identify deceptive or misleading job postings. The study analyzes more than twenty distinct features extracted from five major categories, including evaluation of job description text, examination of company profiles, analysis of job requirements, verification of location details, and assessment of salary information.The developed model effectively detects suspicious patterns commonly associated with fraudulent job listings, such as poor grammatical structure, unrealistic salary offers, and unverifiable company information. To enhance detection performance, the system incorporates several data preparation techniques, including data preprocessing, outlier detection methods, statistical validation techniques, and handling class imbalance in the dataset. Furthermore, the research evaluates the model using appropriate testing strategies and statistical performance metrics, including measures such as precision, recall, and overall accuracy. The results demonstrate that the proposed framework provides a practical and efficient solution for identifying fraudulent job postings. By improving the reliability of job listings, this system can contribute to enhancing the security, transparency, and trustworthiness of online employment platforms.

Keywords : Fraud Detection, Feature Engineering, Predictive Analytics, Random Forest, XGBoost

Author : Dr. E.RAVI,KOMPELLI SHIVAJI, RAGHAVENDRA REDDY,GUNREDDY MEGHANA,KAMMAMPATI SWETHA,BATKA VARUN

Title : Deceptive Recruitment Detection using Machine Learning

Volume/Issue : 2026;03(03)

Page No : 136-142

This paper presents a Real-Time Student Attendance System based on Face Recognition using Artificial Intelligence (AI) to automate attendance management in educational institutions. The proposed system employs advanced deep learning techniques, particularly Convolutional Neural Networks (CNNs), to perform accurate face detection and recognition. It combines facial recognition algorithms with a live video stream, allowing the system to automatically identify students as they enter the classroom environment. Initially, the system detects faces using a reliable face detection method, after which the recognized facial features are compared with previously stored student records in the database. Once a match is confirmed, the student’s attendance is recorded automatically, providing a fast and efficient solution. The integration of AI significantly improves the system’s performance, enabling high recognition accuracy even under different lighting conditions or in situations where slight facial occlusion occurs. In addition, the system incorporates security mechanisms to prevent impersonation or fraudulent attendance marking. Designed to be easily integrated into existing institutional infrastructure, the proposed framework provides a dependable and effective alternative to conventional manual attendance systems. By automating the process, the system reduces time consumption, minimizes human errors, and improves overall transparency in attendance management.

Keywords : Face Recognition Attendance System

Author : S. VASUDEVA CHARY,CHANDRALA SHIVA PRIYA,GUNJI THILAK,CHEPURI SHIVA SAI,DASARI ANIL KUMAR

Title : An automated student real time attendance using face recognition

Volume/Issue : 2026;03(03)

Page No : 143-148

Online Social Networks (OSNs) provide powerful platforms for sharing information, following current events, promoting products, and connecting with people across the world. As a result, millions of users actively participate in social media platforms such as Instagram, Twitter, and Facebook. Despite these advantages, it is often challenging to determine whether a particular social media account genuinely belongs to a real individual or an organization. The creation of fake or malicious accounts allows harmful or misleading content to spread quickly across these networks. Consequently, detecting and predicting fake accounts has become an important research problem in the field of social network security. In this study, machine learning techniques are applied to identify fake accounts on social media platforms. The research also evaluates the performance of different activation functions within Artificial Neural Network (ANN) models. Experimental findings indicate that machine learning approaches are effective in detecting suspicious or fraudulent accounts with satisfactory accuracy. Additionally, the selection of different activation functions in various layers of the neural network significantly influences the classification results. Previous studies in the literature have explored several classification techniques for identifying fake profiles and spammers in online social networks. However, to the best of our knowledge, only limited research has specifically examined the classification of fake accounts using ANN models with multiple activation functions.

Keywords :

Author : M. SHIVARANJANI,DONGALA SWATHI,CHAVABATHINA SHREYA,KANKANALA VENUVARDHAN REDDY,DANDU SAI BHAGEERATH REDDY

Title : Spammer detection and fake user identification on social networks explaining details

Volume/Issue : 2026;03(03)

Page No : 149-158

Understanding customer buying behaviour is a key to knowing how people select and buy products. This research, titled “A Study on Customer Buying Behaviour with Special Reference to TVS Ravi Automobiles ,” aims to explore the factors that affect customers when they purchase TVS two-wheelers and to assess their level of satisfaction. The study examines aspects such as price, mileage, quality, brand image, and after sales service. Primary data was collected through a structured questionnaire from customers of TVS Ravi Automobiles. Secondary data was gathered from books, company records, and websites. The collected data was analysed using simple statistical tool such as percentages, tables, and charts. The finding reveal that mileage, brand reputation, and service quality play a significant role in influencing customers’ purchase decisions. The study also shows that most customers are satisfied with TVS vehicles and the services provided by the dealer. The research offers useful suggestions to improve customers satisfaction and strengthen dealer -customers relationships.

Keywords : Customer Buying Behaviour, TVS two -wheelers, Customer Satisfaction, Brand Image, Dealer services

Author : K. Suma Harshini, S. Fakrulla

Title : A STUDY ON CUSTOMER BUYING BEHAVIOUR WITH SPECIAL REFERENCE TO TVS RAVI AUTOMOBILES, BADVEL

Volume/Issue : 2026;03(03)

Page No : 159-168

Statistical evidence indicates a steady rise in violence involving firearms, making it increasingly difficult for law enforcement agencies to respond promptly. Certain regions, particularly those with relaxed gun regulations, report higher incidences of crimes involving handguns. Ensuring public safety therefore requires the early detection of such threats. One effective approach is the use of surveillance systems capable of identifying dangerous weapons like handguns from video footage, thereby helping to prevent potential incidents. However, many existing surveillance systems still rely heavily on manual monitoring and control.To address this limitation, the proposed system enables automatic weapon detection in video streams for monitoring and security purposes. The system employs the YOLOv3 (You Only Look Once) algorithm to achieve real-time detection of weapons. Earlier approaches, such as R-CNN, require multiple region-based evaluations and involve high computational complexity, making them less suitable for real-time applications. These methods process different regions of an image separately, leading to increased processing time. In contrast, the YOLO architecture processes the entire image in a single pass through the neural network, significantly improving speed and efficiency. Due to its high processing speed and accuracy, YOLOv3 is well-suited for real-time video analysis. When a weapon is detected, the system immediately generates alerts to notify authorities, enabling timely intervention. By facilitating rapid detection and response, the proposed system contributes to preventing violent incidents before they occur and enhances overall public safety.

Keywords : Weapon detection, Surveillance system, YOLOV3

Author : G. UMA MAHESWARI,VELETI V N V PRANEETH,M PRASHANTH,MACHARLA SRI LAVAN,SHARIYA MEHROZ

Title : Weapon Detection With Surveillance With YOLO

Volume/Issue : 2026;03(03)

Page No : 169-179

The rapid growth of smartphone usage, driven by affordability and the digitalization of services, has introduced significant security challenges. Among these, malware threats have become a major concern, particularly on the Android platform, where the proliferation of malicious and fraudulent applications has increased substantially. As Android devices continue to gain popularity, malware developers consistently create new threats, compromising system integrity and user privacy. This study aims to apply Machine Learning (ML) techniques for effective Android malware detection. A comprehensive detection framework is proposed, incorporating six ML models for classifying different types of malware, including Decision Trees, Support Vector Machines (SVM), Naive Bayes, Random Forests, K-Nearest Neighbors (KNN), and Ensemble Methods such as the Extra Trees Classifier. The framework is evaluated using the CICMalAnal2017 dataset, which includes diverse malware categories such as adware, ransomware, and scareware. To enhance model performance, multiple feature selection techniques are employed, including Feature Correlation, Random Forest Importance, Chi-Square Test, and Information Gain. These methods help identify the most relevant features for accurate classification. The performance of various ML algorithms is analyzed and compared to determine the most effective approach for malware detection. Furthermore, the study highlights the importance of using ML-based techniques to detect vulnerabilities at the source code level, as implementing security measures after application deployment can be more challenging. Overall, this research contributes to a deeper understanding of Android malware detection and provides insights into potential future directions for improving mobile security systems.

Keywords : Malware detection; code vulnerability; machine learning; Android security

Author : B. PAVAN KUMAR,POTTABATHULA KEERTHY,TEEGALA TRISHANK,ONTEDDU SANTOSH,VALIMINETI KRISHNA CHAITHANYA

Title : Automated Android Malware Detection

Volume/Issue : 2026;03(03)

Page No : 180-188

Ensuring patient safety is a critical concern in healthcare, where errors or adverse events can have severe consequences. Traditional predictive models often provide limited insight into their decisionmaking processes, making it challenging for healthcare professionals to trust and act on their recommendations. Explainable Artificial Intelligence (XAI) offers a solution by providing transparent and interpretable models that highlight the reasoning behind predictions. This paper presents an XAI-based approach to enhance patient safety by analyzing clinical data to predict potential risks such as medication errors, adverse drug reactions, or hospital-acquired infections. By combining machine learning algorithms with interpretability techniques like SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations), the system not only predicts safety risks but also explains the contributing factors for each prediction. Experimental results on healthcare datasets demonstrate that the proposed system achieves high predictive accuracy while providing clear, actionable explanations for clinicians. This transparency improves trust, facilitates timely intervention, and supports informed decision-making, ultimately enhancing patient safety and care quality.

Keywords : Explainable AI, Patient Safety, Machine Learning, Deep Learning, Clinical Decision Support, Model Interpretability, Risk Prediction, Medical Error Prevention, Feature Importance, Trustworthy AI, Healthcare Analytics.

Author : P. HIMA BINDU,MUNDRAI HIMABINDHU,M SHARANYA,MOHAMMAD SOHEL,TUMMIDI ASHWIDH

Title : Explainable AI For Patient Safety

Volume/Issue : 2026;03(03)

Page No : 189-194

In transformer-less solar photovoltaic (PV) systems, high-frequency capacitance between system components can create unintended pathways that lead to leakage current flow. These leakage currents negatively impact system performance, causing electromagnetic interference and reducing the lifespan of PV panels. To address this issue, this study proposes a passive filter designed specifically to suppress leakage currents during PV array operation. An analytical optimization approach based on frequency-domain analysis is used to determine optimal filter parameters, ensuring efficient performance. Unlike conventional techniques, this method does not require additional semiconductor switches, thereby simplifying the system design and improving reliability. An adaptive control mechanism is incorporated to extract fundamental load current components and effectively reduce harmonic distortion. Simulation results demonstrate that the proposed filter maintains leakage current levels within the limits specified by DIN and NB/T grid standards. Furthermore, comparative analysis highlights the practicality and superior performance of the approach. Real-time Hardware-in-the-Loop (HIL) testing under various operating conditions confirms that the grid current Total Harmonic Distortion (THD) remains below 5%, complying with IEEE 519 and IEC 61727 standards. Overall, the proposed system offers an efficient and reliable solution for improving the performance and longevity of transformer-less PV systems.

Keywords : DC-DC converter, DC-AC converter, DSTATCOM, grid-tied solar array, harmonic analysis, MPPT, power quality, solar power generation, voltage control, voltage source converter.

Author : T.VENU GOPAL, R.SHIVA, J.PAVAN, T.YASHWANTH, R.ANIL

Title : LEAKAGE CURRENT MITIGATION TECHINIQUE IN SOLAR PV ARRAY USING PASSIVE FILTERS.

Volume/Issue : 2026;03(03)

Page No : 195-201

Auscultation of heart and lung sounds remains a fundamental diagnostic tool, yet traditional manual interpretation is often subjective and prone to inconsistency. This study introduces a novel, dualcategory cardio-respiratory dataset captured via digital stethoscopes from a high-fidelity clinical manikin, featuring separate and mixed acoustic recordings—the first of its kind to provide synchronized multi-organ sound profiles. Addressing the rising burden of respiratory and cardiovascular diseases in India, we propose an automated classification framework that leverages a multi-stage feature engineering and deep learning pipeline. Acoustic features, including MelFrequency Cepstral Coefficients (MFCC), Chroma, and Mel-spectrograms, were extracted and evaluated across multiple classifiers, including Quadrature Discriminant Analysis (QDA) and Gradient Boosting (GBC). To capture both local spectral patterns and long-range temporal dependencies, a hybrid Bi-directional Convolutional Neural Network (BiCNN) integrated with a Bidirectional Gated Recurrent Unit (BiGRU) was developed. This architecture enables simultaneous dual classification of heart and lung sound types from a single mixed audio input. Experimental results, validated through accuracy, F1-score, and ROC-AUC metrics, demonstrate that the BiCNNBiGRU model significantly enhances diagnostic precision, offering a robust, non-invasive screening tool for early clinical intervention

Keywords : Cardio-respiratory auscultation, Digital stethoscope, Deep learning in healthcare, Multilabel classification, Bi-directional CNN, Bi-GRU, MFCC.

Author : S. Swapna, Rendedla Supritha, Vemula Deepika, Sardar Avtaar Singh Lohia, Vikram Enugula

Title : An AI-Based Diagnostic Framework for Cardiopulmonary Health Monitoring from Clinical Dataset Analysis

Volume/Issue : 2026;03(03)

Page No : 202-210

The rapid expansion of web-based applications has intensified the need for robust authentication systems to protect user data from escalating cyber threats, particularly SQL injection (SQLi) attacks. This research focuses on developing a Django-based web application that integrates both vulnerable and secured login mechanisms to demonstrate essential secure coding practices. The core problem addresses insecure authentication implementations where improper input handling allows attackers to manipulate database queries. Many traditional systems prioritize functionality over security, utilizing direct query concatenation and minimal validation, which enables unauthorized access and compromises data integrity. To address these vulnerabilities, the proposed system utilizes the Django framework paired with a MySQL database for user registration and authentication. A dual-module architecture is employed: a vulnerable module designed to illustrate SQLi risks, and a secured module that implements rigorous input validation and pattern-based attack detection to identify suspicious characters in real-time. This structure facilitates a comparative analysis between insecure and robust implementations. The significance of this project lies in its ability to bridge the gap between theoretical security and practical application. By showcasing real-world vulnerabilities alongside their respective protective mechanisms, the system enhances the understanding of secure software development lifecycles (SDLC). Ultimately, this research provides a functional foundation for building resilient authentication frameworks, improving system reliability, and promoting awareness of web security standards in modern software engineering.

Keywords : SQLi attack, secure authentication, cybersecurity awareness, Django web framework

Author : Pallavi Gudimilla, Manugonda Kalyani, Koppula Bhargavi Sree, Manchala Sandeep, Palangi Ajay Winsent

Title : Exploit-Aware Security Framework for Real-Time SQL Injection Defense in Login Systems

Volume/Issue : 2026;03(03)

Page No : 211-221

Pest identification is a challenging task in the agricultural sector, as accurate and timely detection of pests is essential for effective pest control and crop protection. Conventional approaches to pest detection, such as entomological knowledge and manual examination, take a lot of time and are prone to human mistakes. The research presents a novel solution designed to automate and enhance pest management in agriculture. The system directly addresses the significant drawbacks of the traditional manual identification method, which is time-consuming, labour-intensive, and reliant on scarce human expertise. Our proposed system provides a fast, accurate, and accessible alternative by leveraging an advanced, hybrid machine learning pipeline integrates deep learning convolutional neural networks (CNN), with Linear Discriminant Analysis (LDA) and Categorical Boosting (CatBoost), here after referred as hybrid neural boosting. The core of the system is a powerful CNN that acts as a sophisticated feature extractor, automatically learning to identify complex visual patterns in insect images. The high-dimensional features from the CNN are then fed into K- nearest neighbours (KNN), eXtreme gradient boosting (XGboost) and proposed LDA with Catboost, a highly accurate boosting algorithm that makes the final classification. The system is designed for practical, real-world application, allowing users to upload a new insect image and receive an instant, on-device prediction. By integrating these robust models, the project not only achieves high classification accuracy but also ensures computational efficiency, making it suitable for deployment on resourceconstrained edge devices without a constant internet connection. The result is a proactive, scalable, and user-friendly tool that empowers farmers with the ability to perform timely and targeted pest control, thereby contributing to sustainable agriculture and improved crop yields.

Keywords : Pest identification, Agricultural sector, Deep learning, Machine Learning, Ensemble learning models.

Author : B. Maheshwari, Uppu Srujana Sri, Polsani Saipriya, Shaik Imtiyaz, Shaik Yakub Pasha

Title : Edge-Compatible Hybrid Deep Learning and Boosting Model for Automated Agricultural Pest Identification

Volume/Issue : 2026;03(03)

Page No : 222-232

Otitis media includes several common inflammatory conditions of the middle ear that can have severe complications if left untreated. Correctly identifying otitis media can be difficult and a screening system supported by machine learning would be valuable for this prevalent disease. The study presents a deep learning-based automated framework for diagnosing otitis media from otoscopic images, addressing the increasing global burden of ear infections, particularly among children, where the disease remains a leading cause of hearing impairment and frequent ENT consultations. Although digital otoscopic devices generate large volumes of clinical images, diagnosis still predominantly relies on manual visual inspection by otolaryngologists, making it subjective, time-consuming, and prone to inter-observer variability, delayed detection, and fatigue-related errors, especially in resource-limited settings. Conventional assessment involves identifying conditions such as Acute Otitis Media, Chronic Otitis Media, Myringosclerosis, Cerumen Impaction, and Normal cases through expert interpretation, which lacks standardization and scalability. To overcome these challenges, the proposed system integrates image preprocessing, deep feature extraction using DenseNet121, and classification using machine learning models including Nearest Centroid Classifier (NCC), K-Nearest Neighbors (KNN), and eXtreme Gradient Boosting (XGBoost), and proposed calibrated perceptron with Dense neural network (DNN) also called as ensemble Voting model to enhance predictive performance. The dataset is divided into training and testing subsets, and the models are evaluated using accuracy, precision, recall, F1-score, confusion matrix, ROC curves, and AUC metrics. Experimental results demonstrate that the Voting-based ensemble model achieves superior performance compared to individual classifiers, providing a reliable, efficient, and scalable solution for automated otitis media detection and clinical decision support.

Keywords : Early Otitis, Otoscopic imaging, Inflammatory conditions, Chronic otitis media, Conventional assessment.

Author : Zareena Begum, Martha Navya Sri, Md Afrin, Kotte Manesh, Manugonda Muneeshwari

Title : Deep Hybrid Medical Vision System for Early Otitis Media Diagnosis using Otoscopic Imaging

Volume/Issue : 2026;03(03)

Page No : 233-244