Browse Items (3095 total)
Sort by:
-
Utilizing Transforming Portfolio Management Through Automation Using Advanced Deep Reinforcement Learning Algorithms for Optimized Investment Strategies
This paper focuses on the future possibility of enhancing the applications of DRL in autonomously managing a portfolio for better investment plans. Having used past financial data and a highly developed case of DRL, the proposed system shows better performance compared to conventional investment strategies and indices. This process includes data gathering from the financial databases, the steps of preprocessing and feature extraction, and the use of the DQN structure. After that, the system's training and validation are done by a finite portion of real-world data and a large number of synthesized data to improve stability. The result shows that the new method achieves superior cumulative return, Sharpe ratio, maximum drawdown, and annualized volatility; therefore, it suggests that the proposed system can flexibly predict the fluctuating stock market trends and make appropriate investment decisions. Thus, the present research adds importance to the use of DRL in improving return potential and risk management in portfolio management. Thus, this study adds to the existing literature and practice by allowing for the automation of the optimization and testing for investment solutions at a larger scale, while opening up opportunities for future developments in the application of financial technology and investment tools. 2025 IEEE. -
Synergistic Hybrid Segmentation for Handwritten Kannada Word Recognition Addressing Deep Learning Challenges
The handwritten Kannada script has an intricate aksharas that are formed by combining consonants, vowels, and ottus. These complex combinations pose significant hurdles for automated text segmentation. The inherent diversity in handwriting styles, coupled with prevalent character overlap, multi-touch connections, varied curve structures such as upper open curve(OC), upper closed curve (CC), and the highly condensed nature of Ottaksharas, routinely blurs character boundaries, leading to severe segmentation errors that propagate and compromise overall recognition accuracy. A hybrid approach that customizes adaptive traditional methods like vertical pixel count, to identify true character gaps in handwritten Kannada characters could effectively manage character overlap, or segment multi-touch characters or Ottaksharas. This pre-processing stage can allow subsequent deep learning models to recognize this segmented character. This will avoid significant hurdles: immense data requirements for pixel-level annotations, high computational costs for dense prediction, and significant architectural complexities for precise boundary delineation and handling connectivity. Given these constraints, particularly with less-resource language like Kannada, scaling deep learning models will lead to ever erroneous recognition. This paper argues that modified traditional approaches, by directly embedding customized knowledge and leveraging targeted feature engineering, can offer a computationally efficient and data-lean alternative. This strategy enables more robust segmentation for complex Kannada characters, providing a practical pathway for automated handwritten text processing in such linguistic domains. 2025 IEEE. -
Fraud Prevention in Banking: Machine Learning-driven Approaches for Detecting Payment Anomalies
The fast-paced development of digital banking has brought with it new convenience but also tremendous challenges in maintaining transaction security. Banks are confronted with mounting threats from malicious activities like identity theft, account takeover, and unauthorized access, which can lead to huge financial losses and loss of customer confidence. This study investigates the formulation of a cybersecurity framework for fraud prevention in banking through machine learning algorithms. A transactional real-world dataset of 200,000 instances from LOL Bank Pvt. Ltd. was used to construct and evaluate predictive models. Preprocessing included categorical encoding, temporal feature engineering, and synthetic minority oversampling (SMOTE) for class imbalance handling. Three machine learning classifiers - Logistic Regression, Random Forest, and XGBoost - have been compared using measures of accuracy, precision, recall, F1-score, and ROC-AUC. Results show that ensemble models significantly outperformed logistic regression by a wide margin, with Random Forest and XGBoost both achieving over 91% accuracy and very good discrimination power. The study emphasizes how well machine learning-based systems detect theft in real time and outlines avenues for future research to enhance detection using adaptive and interpretable AI models. 2025 IEEE. -
Enhancing Diagnostic Precision in Lung Cancer Detection Using Smote-Based Balancing Techniques
Worldwide, Lung cancer is the primary cause of death from cancer, and chances of surviving are considerably raised by early detection. While traditional diagnostic approaches heavily rely on imaging and specialized infrastructure, they often fail to serve low-resource or early screening environments. In this work, based on deep learning, lightweight framework for detecting lung cancer from structured survey data is presented. The research tackles the prevalent the problem of class disparity using the Synthetic Minority Over-sampling Technique (SMOTE), enhancing the sensitivity of predictive models. A comparative evaluation was conducted across six models Logistic Regression, SVM, KNN, Naive Bayes, Random Forest, and XGBoost. Among these, Random Forest and XGBoost achieved 95% accuracy, 0.98 recall, and ROC-AUC scores of 0.9943 and 0.9835 respectively. The proposed hybrid ensemble model (Random Forest + XGBoost) outperformed all with 96% accuracy, 0.95 precision, 0.98 recall, and a ROC-AUC score of 0.9961. These findings demonstrate that the hybrid strategy is effective in providing high diagnostic precision using clinical survey data that is not imaging. 2025 IEEE. -
An Integrated Approach to Green Cloud Solutions for Energy-Efficient Sustainable IT and Carbon Footprint Reduction
Cloud computing has become a very important part in everyday life, but this has also made a lot of carbon footprint because of the energy consumption in the data centers. The pandemic had affected these emissions, and they quickly came back, which has shown the requirement for sustainable solutions which will help in fighting the increase in carbon footprint. For these problems, the green computing technology will give probable solutions by promoting the technology that would be responsible enough to decrease these effects of harming environment. It will have techniques like smarter system designs, operations that are energy efficient, and smart techniques for optimization. This study explores how the above set principles can reduce the overall digital carbon footprint and help to create economically viable businesses. This approach provides a forward path for technology progress and profitability aligning with the environment sustainability which is a necessary component for business longevity. 2025 IEEE. -
MCCLDP: Multi Class Cotton Leaf Diseases Prediction and Classification using Deep Learning Model
Cotton plant disease detection is critical for sustainable agriculture and reducing crop losses. This paper proposes a novel Multi-Stream Attention-Guided Hybrid CNN (MAH-CNN) for accurate classification of cotton leaf diseases. The model leverages pre-trained ResNet152v2 and DenseNet-121 backbones for hierarchical feature extraction, complemented by a shallow CNN for localized texture analysis. A spatial attention mechanism enhances focus on disease-relevant regions, mitigating background noise. Features from the global and local streams are fused and passed through a lightweight classification head. The model achieves superior performance in terms of accuracy 97.32%, F1 score 98%, and specificity 100% on benchmark datasets which are available in open access, outperforming existing state-of-the-art methods. The integration of Grad-CAM provides interpretability, fostering trust in automated disease detection systems. 2025 IEEE. -
Self-Organizing Microservice Composition for IoT-Enabled Neonatal Monitoring
The fast-paced growth of the Internet of Things (IoT) has revolutionized several industries, including the delivery of healthcare applications, due to its ability to support realtime and automated applications. However, several key challenges are associated with the dynamic and complex nature of IoT environments, such as scalability, increased system complexity, and network instability. This paper presents a self-organizing microservice Composition Model created especially for healthcare systems using IoT to address these issues. The Model employs decentralized, local interactions inspired by bio-inspired mechanisms. These mechanisms enable the system to autonomously form complex service composites with minimal human intervention through emergent behaviour. The model leverages an ant agent to optimize bio-inspired mechanisms and a deep learning agent for autonomous service choreography, enhancing the system's flexibility, adaptability, and overall performance. A continuous neonatal monitoring case study illustrates how the model works in composing services in a self-organizing manner. This research contributes to the advancement of smart healthcare infrastructures by proposing an innovative framework that adapts to the evolving needs of IoT-driven healthcare applications. 2025 IEEE. -
Self-Organizing Microservice Composition for IoT-Enabled Neonatal Monitoring
The fast-paced growth of the Internet of Things (IoT) has revolutionized several industries, including the delivery of healthcare applications, due to its ability to support realtime and automated applications. However, several key challenges are associated with the dynamic and complex nature of IoT environments, such as scalability, increased system complexity, and network instability. This paper presents a self-organizing microservice Composition Model created especially for healthcare systems using IoT to address these issues. The Model employs decentralized, local interactions inspired by bio-inspired mechanisms. These mechanisms enable the system to autonomously form complex service composites with minimal human intervention through emergent behaviour. The model leverages an ant agent to optimize bio-inspired mechanisms and a deep learning agent for autonomous service choreography, enhancing the system's flexibility, adaptability, and overall performance. A continuous neonatal monitoring case study illustrates how the model works in composing services in a self-organizing manner. This research contributes to the advancement of smart healthcare infrastructures by proposing an innovative framework that adapts to the evolving needs of IoT-driven healthcare applications. 2025 IEEE. -
Price Minds: AI-Driven Insights, Recommendations and Dynamic Pricing
This research aims to enhance e-commerce systems by leveraging customer behavior analysis, dynamic pricing, and personalized recommendations. With the increasing demand for tailored shopping experiences and competitive pricing, businesses require adaptive solutions. The study integrates synthetic and real-time customer data to identify purchasing patterns and segment customers effectively. Dynamic pricing strategies are applied to optimize revenue while maintaining customer satisfaction. A unified framework combines clustering techniques, real-time data streams, and decision-making models to deliver actionable insights for business operations. The proposed system dynamically adjusts pricing and recommends products based on individual customer preferences and behavior. The approach addresses the growing need for intelligent systems that adapt to market trends and consumer demands. Results demonstrate improved operational efficiency, better customer engagement, and enhanced profitability. This work highlights the importance of real-time analytics and intelligent pricing mechanisms in advancing e-commerce and creating competitive advantages in rapidly evolving markets. 2025 IEEE. -
Levelling Up for Nature: Pathways of Engagement with Green Messages in Games
Digital games have become potent platforms for environmental messaging, capable of influencing beliefs, attitudes, and behaviours. This paper explores how players engage with green content embedded within game environments, using mixed-method data from the GREAT Project with more than 181,000 anonymised gameplay sessions. This research analysed cognitive, affective and behavioural responses to in-game environmental messaging and identified key pathways that lead to sustained pro-environmental actions using PCA and K-Means clustering analysis, by which varying types of user were identified from casual to deeply engaged players. The engagement questions were grouped into three categories: narrative frame, behavioural pledges, and social reflection. Our findings showed that 68% of the players participated in at least one green-themed pledge, which indicates a strong sensitivity to environmental content, informing a multilayered engagement model that supports both player experience and environmental literacy known as a 5-pathway green engagement model (5-PGEM), supporting the hypothesis that green messages, when embedded in games, can significantly shape eco-friendly actions. 2025 IEEE. -
An Intelligent Framework for Evaluating Handwritten Responses: Integrating Bloom's Taxonomy with Adaptive Assessment
Traditional manual grading of descriptive-type answer scripts is inefficient and laborious, while existing technologies rely on strict keyword matching and cosine similarity, which fail to capture the semantic meaning and argumentative quality. This paper proposes a multi-layered intelligent framework for evaluating handwritten descriptive answer scripts by integrating Revised Bloom's Taxonomy with adaptive assessment methods. The system consists of a multi-dimensional evaluation strategy comprising four valuation metrics, namely, content relevance, coherence, depth, and argumentation quality. When compared with conventional methods of keyword matching or computing the cosine similarity, this proposed framework evaluates the semantic meaning and argumentative structure while adapting to varying response styles and contextual differences for personalised assessment. 2025 IEEE. -
MobileNetV3-Based Fine-Tuned Facial Emotion Recognition with Targeted Class Balancing
Facial emotion recognition (FER) is a pillar of affective computing and augmented human computer interaction, but has been stymied by the problem of class imbalance and lack of prevalence of subtle emotional differences. This paper presents a lightweight FER framework based on the MobileNetV3 architecture with a fine-tuned and weighted dataset that applies class balance and class weighting as strategies that optimized the three-class classification of three discrete emotions Angry, Happy, and Sad. The characteristics of the dataset were assembled comprising a total of 7,305 labelled facial images, based on the KDEF, Kaggle, and Face Expression Dataset hence inheriting the heterogeneity of subjects and imaging conditions. The pre-processing of all of the images carried out as the RGB input and after resizing (224 x 224 pixels) a massive data augmentation done to encourage generalization. Transfer learning in the training pipeline is done through progressive unfreezing and the weight of the loss on the minority classes (Angry and Sad) are boosted to improve the performance of detection. The achieved model resulted in an accuracy of 87% on the test set, and had equal accuracy in preciseness, recall, and F1-scores over all emotion types. Extended error analysis revealed that the majority of cases that were misclassified fell between the categories Angry and Sad because they were mistaken due to combining visual cues. Even then, the performance showed stability despite the variable lighting as well as in variable positional context. In Comparison, MobileNetV3 outperforms state-of-art-lightweight models with respect to accuracy and computation of similar computational complexity. 2025 IEEE. -
Contextual Embedding Fusion for Quick-Commerce Reviews: Rating, 5-Class Classification, and Sentiment Analysis Across Classical ML, DL, and Transformer Models
Quick-commerce platforms depend on accurate reading of customer feedback to steer pricing and service quality. We study a dataset of delivery-agent reviews, and build two rating predictors and a text-only classifier suite. On the rating regression task (1-5), we benchmark seven models (Ridge, Lasso, ElasticNet, SVR-RBF, Gradient Boosting, Random Forest, XGBoost). The best test performance is achieved by RandomForestRegressor with MAE = 0.527, RMSE = 0.973, and R2 = 0.555; cross-validation places XGBoost and Gradient Boosting close behind. For 5-class rating classification, we compare seven TF-IDF baselines (LogReg, LinearSVC, two SGD variants, RidgeClassifier, ComplementNB, XGBClassifier) and find SGD (hinge) strongest (Accuracy = 0.817, Macro-F1 = 0.389), yet mid-rating classes (2-4) remain difficult. To address label ordinality, we fine-tune DistilBERT with ordinal heads: CORAL and CORN. These achieve Accuracy = 0.837/0.819 and Macro-F1 = 0.439/0.441, improving Macro-F1 by +5-6 points over TF-IDF baselines, mainly by rescuing classes 2-4. We further map regression outputs to 5 classes via optimized cutpoints, raising Macro-F1 from 0.329 (rounding) to 0.380. Paired, fold-wise comparisons using Wilcoxon signed-rank and bootstrap 95% CIs show consistent gains; with 5 folds the discretized p = 0.0625 indicates all folds favor the better model, suggesting statistical significance with more folds. Overall, ordinal transformers and learned thresholds provide measurable, reproducible improvements for short, noisy e-commerce reviews, delivering higher fidelity on mid-range ratings while preserving high accuracy. 2025 IEEE. -
Gaming for the Planet: Exploring Belief, Motivation, and Behaviour in Climate-Conscious Play
The emergence of digital games as an impactful medium for engaging audience from diverse background and clime cannot be overemphasised in today's world of sophisticated technologies and high-end devices. This study explored the potential capabilities of digital games in influencing beliefs, motivations, and behaviours that have connections with climate. The main purpose was to understand the psychological and structural game design elements that promotes measurable ecological impact. Case studies from the EU-funded GREAT Project, such as Play2Act, and broader literature on gamification for environmental engagement. Data from a global in-game poll (n > 180,000) was analysed to gain insights into how green messaging within popular games translates into real-world behavioural changes. Findings from the study shows that 65% of players reported increased awareness of climate issues. Also, results indicated that 40% of the players adopted at least one new eco-friendly behaviour after gameplay. Finally, it was found that specific game design elements like reward systems and narrative integration. The study concluded that mainstream digital games could serve as scalable tools for environmental engagement, which can bridge the gap between environmental psychology, game studies, and design research. We recommend that future research should focus on refining these design frameworks and integrate them into policy to maximise the ecological impact of digital games. integration. 2025 IEEE. -
Design of a Multi Camera Enabled Scrutinizing Framework for Smart Cities
This research paper presents a multi-camera surveillance system tailored for the demands of smart cities. By integrating edge computing, the system decentralizes processing, reducing latency and alleviating network bandwidth strain. The architecture includes layers for data collection, edge processing, centralized storage, AI-driven analysis, synchronization, and user visualization. Cameras capture and preprocess data locally to identify anomalies and minimize unnecessary transmission. AI algorithms handle tasks like object tracking, behavior analysis, and event detection with precision. Synchronization ensures seamless temporal alignment across video streams for accurate event reconstruction. User-friendly dashboards provide actionable insights for urban planning and public safety. By leveraging edge computing, AI, and robust synchronization, this system addresses scalability, latency, and privacy concerns, offering enhanced safety, optimized traffic flow, and better urban planning. 2025 IEEE. -
An Optimized Convolutional Neural Network Model for Real-Time Object Detection in Drones
The capacity of drones to perform item detection in actual-time is crucial for applications inclusive of surveillance, seek and rescue, and environmental tracking. This look at investigates how convolutional neural networks (CNNs) can beautify object detection in aerial imagery by enhancing both accuracy and speed. CNNs excel at extracting spatial info, permitting drones to apprehend objects even in relatively complicated environments. by adopting light-weight CNN architectures and optimization strategies, we acquire advanced performance with minimum computational requirements, ensuring green operation on embedded drone platforms. Our findings verify that CNN-based fashions considerably decorate detection accuracy and responsiveness, allowing the improvement of smarter and more self reliant drones. 2025 IEEE. -
An Optimized Convolutional Neural Network Model for Real-Time Object Detection in Drones
The capacity of drones to perform item detection in actual-time is crucial for applications inclusive of surveillance, seek and rescue, and environmental tracking. This look at investigates how convolutional neural networks (CNNs) can beautify object detection in aerial imagery by enhancing both accuracy and speed. CNNs excel at extracting spatial info, permitting drones to apprehend objects even in relatively complicated environments. by adopting light-weight CNN architectures and optimization strategies, we acquire advanced performance with minimum computational requirements, ensuring green operation on embedded drone platforms. Our findings verify that CNN-based fashions considerably decorate detection accuracy and responsiveness, allowing the improvement of smarter and more self reliant drones. 2025 IEEE. -
An Intelligent Cognitive Framework for Crime Prediction in Smart Cities using Video Mining
Booming development in cities with dense population have led to urban policing and public safety emerging as urgent concerns in city environments.current monitoring practices including CCTV'S and other IOT sensors generate a vast amount of data ,thus making them inadequate for the task. However a combination of video mining,computer vision,artificial intelligence and data mining techniques,do offer us a better framework for monitoring and real-time detection of crime in Smart city"s environment. This paper proposes an intelligent and Cognitive framework for prediction of crime. by combining various advanced modals such as YOLO (You took Only Look Once) for detecting objects, 3D Convolutional Neural Networks (CNN) for recognizing actions, deep SORT for tracking multiple objects, One-class SVM for detecting anomaly and LSTM for behavioral analysis. These modals can organized to function in a coherent system which can be organized to distinguish examine and trail illegal activities such of mugging, robbery, pick pocketing, violence, utilizing available live video feeds. By efficient date processing, and overcoming shortcomings such of limited labeled datasets and real-time feed detection, this framework can provide practical conclusion making tool for law enforcement in urban smart city environments which can enhance urban safety.Besides effective crime detection,this tool compiles with established ethical standards such as upholding privacy and legal compliance. 2025 IEEE. -
Deployment of Smart Surveillance System using Deep Learning to Recognize Cyber-criminals
This paper presents the development of the smart surveillance systems critical in identifying cybercriminals through the use of deep learning.The system utilizes deep learning algorithms for the identification of cybercriminals in physical and cyberspace.The systems apply neural networks to analyze images,video streams and cyber behavior for pattern recognition of suspicious activities and potential threats.Also the system analyze the online activities of users and flow of data within the network for signs of cybercriminals. Various technologies such as convolutional neural networks (CNN), and recurrent neural networks (RNN) are used to distinguish facial features, body language, and unusual online activities. This way, effective security measures are taken to prevent or reduce the impact of cybercriminals in various environments by combining intelligent monitoring systems with future threat prediction. The system is capable of evolving by identifying new criminal patterns to enhance its performance. This means the system is modified and updated as it receives more data, making it effective in multivariate settings such as any institution with financial activities, government networks, and high-security locations. 2025 IEEE. -
Optimizing Sustainable Agriculture Through Customized Crop Management Approach
Sustainable agriculture is essential for actively addressing the dual challenges of global food security and climate variability. This study deploys an intelligent, data-driven approach to tailored crop management, positioning it as a dynamic framework for optimizing cultivation by enhancing soil quality and climate resilience. Recognizing soil quality as a foundational element of sustainable agriculture, this paper highlights its critical role in nutrient cycling, water retention, and organic enrichment. Through strategic interventions such as precision crop rotation, conservation tillage, cover cropping, and organic amendments, this approach maximizes soil porosity, health, and fertility while mitigating environmental degradation. By operationalization tailored crop management as an adaptable and scalable system, this research advances the synergy between agricultural productivity and environmental sustainability. Leveraging AI-driven insights, predictive modeling, and modular frameworks, these strategies empower global efforts toward food security, ecological balance, and climate-adaptive farming solutions. 2025 IEEE.