Browse Items (16481 total)
Sort by:
-
Celestial Image Classification using Ensemble Learning and Vision Transformers
Astronomical image classification plays a crucial role in understanding the universe, but deep learning models often stumble when faced with scarce labeled data. In our work, we address this gap in two key ways: first, by building a richly varied dataset from just 600 Hubble Space Telescope images and, through targeted augmentation, expanding it to 4,500 distinct training examples; and second, by introducing a hybrid learning strategy that marries transformer-driven feature extraction with gradient-boosted decision trees. We used benchmark standard convolutional architectures (ResNet-50, DenseNet-121) alongside the Data-Efficient Image Transformer (DeiT) and two novel hybrids-DeiT-RF and DeiT-XGBoost (DXg). In DXg, DeiT captures complex spatial patterns, an adaptive dimensionality reduction layer hones in on the most informative features, and XGBoost delivers the final classification. This fusion not only boosts accuracy across nebulae, galaxies, and star clusters but also enhances interpretability by revealing which transformer-derived features most influence the model's decisions. 2025 IEEE. -
AI-Driven Lead Scoring: Enhancing Real Estate Decisions with Predictive Analytics
Lead optimization remains underutilized in customer acquisition, with businesses often focusing on new models rather than refining existing processes. Many overlook automation, real-time data, and feedback loops that enhance insight into lead behavior. Automated optimization continuously improves lead scoring by fine-tuning models over time. Current approaches rely on basic lead scoring without real-time data integration or continuous updates. This research focuses on machine learning-driven lead optimization to improve scoring accuracy and personalized communication. We propose an AI-enhanced system that integrates CRM data with predictive models using ensemble techniques like Random Forest and XGBoost. Our approach achieves high accuracy in property hotspot and ROI prediction, with R2 values up to 0.99. However, a 5% uncertainty exists, requiring carefully generated synthetic datasets. This methodology improves lead prioritization, decision-making and data-driven strategies, ultimately increasing conversion rates and revenue growth. 2025 IEEE. -
Genetically Optimized GAT Recommender
In the digital era of business growth in areas such as e-commerce, online food tech, Health tech, Edutech, etc., A significant issue encountered by businesses to customers (B2C) companies is the recommendation of products, particularly for cold start users. The research proposes a recommendation system using Graph Attention(GAT)-based architecture with a genetically optimized clustering algorithm, which solves the general and cold start users problem in a recommendation. In order to solve the general recommendation problems, the proposed architecture is applied to a benchmark dataset from Amazon, with a recommendation accuracy of 99% and for cold start users, the Proposed model will add an extra layer of personalization as a survey will be included during the registration process of the user. The dataset for the survey was collected using Google Forms, and the dataset comprises five primary attributes: name, gender, user type, interested domains, interested products of new users, and recommended products of old users. A clustering analysis was performed on the dataset, and DBscan was used with genetic optimization as it outmatched other clustering algorithms. Then, transfer learning was applied to the survey collected with the same architecture, and it achieved an accuracy of 79.52% with an MSE loss of 0.0664. 2025 IEEE. -
IoT-Enabled Smart Security Surveillance System for Farmland and Livestock Monitoring Using Computer Vision
Agriculture is the main source of income of every country and agriculture plays a critical role in sustaining human civilization by providing food, fuel, and other essential resources. Even though, the conflict between farmers and wildlife remains a significant challenge even today. It is essential to arrange the protection of fields and farms by deterring wild animals and predators without making harm to the wildlife. This study proposes a framework that can detect and classify animals or intruders or any natural calamities. The novelty of the study is deterring the wild animals using innocuous devices and chemical components instead of the harmful electric fences and other methods. Based on the classification it will take preventive measures and will send an alert to the Farmer. The preventive measures are ensuring that will not make any harm to the wild animals. The framework incorporated four major components PIR (Passive Infrared) Sensor, Raspberry Pi Camera, Scarecrow and GSM (Global System for Mobile Communications) Module. PIR Sensor detects an object, Raspberry Pi Camera captures the images and images classify the specific object, later based on the type of the animal the scare tactics will be applied by the scarecrow and if any unusual incident is happening like the presence of an intruder, fire or tornado the alert will send to the Farmer with the help of a GSM Module. 2025 IEEE. -
IoT and Eye Tracking based system for Cerebral Palsy Diagnosis and Assistive Technology
Children with Dyskinetic Cerebral Palsy often suffer significant challenges in interacting with digital devices due to their impaired motor control which hinder the use of traditional input methods such as keyboard and touch screens. The limitations can be overcome by developing assistive technology such as the IOT based system that the paper proposes. The paper proposes an eye-tracking system that uses web-cameras along with computer vision algorithms to detect gaze direction and blink patterns paired with a handheld IOT console that bridges the interaction between the user and the digital interface. The study also proposes ways to assess and evaluate a patient's motor and cognitive functions. The proposed solution keeps in mind the motor limitations of a DCP patient and aims to enhance the accessibility of digital interfaces, providing children with severe motor impairments a more intuitive and inclusive means of interacting with technology. 2025 IEEE. -
Solving Wordle Using Actor-Critic Reinforcement Learning
The popular word-game Wordle poses a difficult sequential decision-making problem with enormous discrete action spaces and partial feedback. In order to solve Wordle as efficiently as possible, this work explores the use of actor-critic reinforcement learning techniques. We develop two actor-critic variations, Vanilla Actor-Critic (AC) and Advantage Actor-Critic (A2C), and formulate Wordle as a Markov Decision Process. Through curriculum training on increasingly larger vocabulary, our method filters out invalid actions and guides learning by combining neural networks and symbolic reasoning. While the AC agent has 42.35 % success with an average guess of 4.85, the A2C agent has a 46.05 % success rate averaging 5.31 guesses per successful game. We show that, especially in the worst-case situations, batch-based A2C performs more robustly than stepbased AC. By effectively scaling from tiny vocabularies (50 words) to the entire Wordle lexicon (14,855 words), our neuro-symbolic technique demonstrates the efficacy of curriculum learning for challenging word games. 2025 IEEE. -
Truth Twisters: Large Language Models Beating Humans at Fake News
Misinformation has become a serious global problem, affecting the process of referendum and decision-making in areas such as politics, healthcare, and social movements. With the rise of advanced artificial intelligence, especially large language models (LLM), the scenario of misinformation building has changed dramatically. These models, which are known to generate coherent and human reactions, can also be used to generate reliable but false or harmful materials. This study examines the dual nature of LLM and highlights its possible misuse to create misinformation that can evade the identity mechanism. The objective of this article is to explain how LLMs can be manipulated through prompt engineering and vocabulary attacks, where adversaries use obfuscated or subtly altered language to bypass content filters and safety guidelines. Despite being fine-tuned for ethical alignment, many LLMs can still be 'jailbroken' - a process by which users modify prompts to elicit inappropriate or restricted outputs. Through a series of controlled experiments, we demonstrate sensitivity to such adverse information of state -of -the -art LLM. These findings create serious concerns about the deployment of LLM in an open-wheel environment. Although these models offer immense possibilities of innovation and productivity, their sensitivity to manipulation outlines the immediate need for strong safety measures. We conclude by discussing moral implications and proposing strategies to reduce abuse, such as better adverse training, strict deployment protocols, and continuous monitoring to balance between safety and innovation in AI. 2025 IEEE. -
Spatiotemporal Forecasting and Environmental Driver Modeling of Marine Microplastic Pollution: an Interpretable Deep Learning Approach for Sustainable Ocean Policy
Marine microplastic contamination presents a significant risk to ocean health, necessitating precise spatiotemporal predictions for effective marine policy development. This study introduces a transparent deep learning model to examine and forecast microplastic levels in global oceans by leveraging historical sampling data, seasonal variations, and climatic factors. A comprehensive global dataset is curated and analyzed, integrating environmental indices such as ENSO, PDO, NAO, and MEI to model the influence of large-scale ocean-atmosphere interactions. Temporal decomposition, Mann-Kendall trend testing, Theil-Sen regression, and seasonal analysis reveal statistically significant monthly and interannual variations in microplastic concentration. Correlations with climate drivers underscore the dynamic environmental control on pollutant distribution. By incorporating interpretable environmental modeling, the proposed framework supports data-driven marine pollution mitigation and policy strategies aligned with UN Sustainable Development Goal 14 (Life Below Water). This work establishes a foundation for future extensions involving LSTM- and Transformer-based time series forecasting combined with SHAP-based explainability for enhanced decision-making. Furthermore, anomaly detection employing Prophet residuals and Isolation Forest reveals sudden increases in pollutants, providing early warning systems for disturbances to marine ecosystems. High-risk areas that need focused regulatory actions are further identified using clustering analysis. All things considered, the model makes it possible to forecast marine plastic pollution in a comprehensive, comprehensible, and scalable manner-a crucial component of sustainable ocean governance. 2025 IEEE. -
Enhancing Network Topology with ONOS, P4 Runtime, and BMV2 Switches
The increasing complexity of modern networks demands advanced solutions for efficient and adaptive topology management. Traditional networking approaches, characterized by their rigid and hardware-centric architectures, often fall short in addressing the dynamic requirements of contemporary networks. This paper introduces a novel approach to network topology management by leveraging the Open Network Operating System (ONOS), P4 Runtime, and BMV2 switches. ONOS, a scalable and distributed SDN controller, provides centralized control and a global view of the network, while P4 Runtime offers a protocol-independent interface to manage programmable data planes. BMV2, a versatile software switch, emulates the behavior of P4-programmable hardware, allowing for the development and testing of custom packet processing pipelines. This research lays the groundwork for future developments in programmable networks by bringing out the potential of combining SDN with P4-based data plane programmability to meet the evolving demands of modern network environments. 2025 IEEE. -
Enhancing Metabolomics Pathway Prediction with Sequential Graph Convolutional Network
Metabolomics is a powerful tool for the understanding of biological systems by analysis of metabolites and their related pathways. Prediction of metabolic pathways is still one of the most challenging tasks because of the complexity of molecular structures and graph-structured metabolomics data. This article presents a robust framework using Graph Convolutional Networks (GCNs) to address the challenges. The methodology proposed includes first preprocessing through metabolite identification by mass spectrometry, and then it utilizes feature extraction through the RDKit library. The objective of the research is aim to metabolic pathway prediction using machine learning algorithm. Complex patterns and relationships are captured from the SMILES representation through the molecular graphs constructed and passed on for the GCN model to learn structured data. ReLU activation functions have been employed within a three-layer sequential GCN architecture that enables it to deliver highly accurate results while ensuring that they are understandable as well. The proposed sequential GCN Model was evaluated on the KEGG dataset with an accuracy of 98.00%, precision of 92.10%, and recall of 93.02%. The performance of these metrics is well beyond traditional approaches such as KNN, ensemble logistic regression, and other GCN based approaches. Thus, this work brings GCN based approaches closer to revolutionizing metabolic pathway prediction and the advancement of the metabolomics field. 2025 IEEE. -
AI-Driven Time Series Models for Rainfall Prediction: A Machine Learning Approach
Rainfall variability plays a critical role in agriculture, water resources, and disaster management in India. This study focuses on the usage of a SARIMAX model to analyze and predict rainfall patterns while accounting for both non-seasonal and seasonal components. The model adequately captures seasonal variations, illustrating the relevance of incorporating seasonal factors into forecasting. However, the non-seasonal components did not considerably improve the models performance. Diagnostic tests demonstrate that the model handles autocorrelation satisfactorily; however, the residuals exhibit irregularities, as evidenced by skewness and large tails. While the model has fair prediction accuracy, the findings show areas for development, particularly in fine-tuning non-seasonal dynamics and eliminating residual abnormalities. This research emphasizes the importance of seasonality in rainfall forecasting and lays the framework for future model improvements. The study provides valuable information on rainfall patterns, supporting better planning and management. 2025 IEEE. -
AI-Driven Time Series Models for Rainfall Prediction: A Machine Learning Approach
Rainfall variability plays a critical role in agriculture, water resources, and disaster management in India. This study focuses on the usage of a SARIMAX model to analyze and predict rainfall patterns while accounting for both non-seasonal and seasonal components. The model adequately captures seasonal variations, illustrating the relevance of incorporating seasonal factors into forecasting. However, the non-seasonal components did not considerably improve the models performance. Diagnostic tests demonstrate that the model handles autocorrelation satisfactorily; however, the residuals exhibit irregularities, as evidenced by skewness and large tails. While the model has fair prediction accuracy, the findings show areas for development, particularly in fine-tuning non-seasonal dynamics and eliminating residual abnormalities. This research emphasizes the importance of seasonality in rainfall forecasting and lays the framework for future model improvements. The study provides valuable information on rainfall patterns, supporting better planning and management. 2025 IEEE. -
Efficient Pathfinding in a Maze to overcome Challenges in Robotics and AI Using Breadth-First Search
Efficient pathfinding in a maze is a key obstacle in robotics, computer science, and artificial intelligence. The article is proposing a strategy using the Breadth-First Search (BFS) algorithm to establish the shortest path for a robot navigating from the top-left to the bottom-right corner of a maze depicted as a two-dimensional grid. The maze comprises open pathways and obstructions, signified by 0 and 1, respectively. The robot's permissible actions include up, down, left, and right, restricted by the boundaries of the grid and the position of obstacles. BFS, an approach well-suited for unweighted graphs, sequentially examines all available routes, ensuring that the first observed path to the goal is the shortest. A visited set removes redundant cell visits, reducing infinite loops and inefficient processing. The algorithm's efficiency is dramatically upgraded by harnessing a queue structure to maintain live routes and their associated steps. This approach assures effectiveness and extensiveness for grid-based navigation problems, making it especially appropriate for real-world robotic applications where minimizing traversal cost is critical. Additionally, the paper discusses the algorithm's execution, complexities, and potential upgrades for larger grids or dynamic environments. Experimental results demonstrate BFS's resilience and efficacy in solving pathfinding challenges in various maze configurations. This work contributes to developing stable navigation techniques, integral to advancing autonomous robotic navigation and related fields. 2025 IEEE. -
Internet of Things Enabled Smart Hand Gesture Virtual Mouse System
This research is aim to focus on IoT based hand gesture model. Mouse is one of the most important input devices of a computer. It works as a pointing device and allows the user to move the pointer as needed by the user. In the early days, a wired mechanical mouse was used for this purpose. In mechanical mouse a ball is fixed underneath the mouse, which rotates as the user moves the mouse. This movement of the ball is used to move the mouse pointer on the screen. Now mostly we use optical mouse which can be wired or wireless. An optical mouse has a high-power laser below it, which takes more than thousand pictures of the surface below the mouse. An image comparator compares the images and sends the signal to move the mouse pointer as the texture of the image changes. Both the types of mouse works based on old technology. As technology leaps to greater heights, the need for simplicity also increases. With the invention of different kind of sensors, microcontrollers and other electronics, we can eliminate the mouse as an input device and instead use our hands to do the work of a mouse. This prototype is an embedded system which runs with the help of an arduino microcontroller. Flex sensors are used to capture the hand gestures. The proposed IoT based hand gesture model is providing high accuracy rate compare to the regular model. The proposed model is analyzed with accuracy level, the average accuracy level of proposed model is more than 90%. 2025 IEEE. -
The Evolution of Cloud Computing: A Study of Aspirational Technologies and Practical Achievements
Cloud computing has transformed the digital landscape, providing scalability, cost efficiency, and seamless access to computing resources. Yet, there is a gap between its theoretical aspirations and real-world achievements because of data privacy concerns, regulatory compliance, vendor lock-in, and performance bottlenecks. This paper critically examines these disparities through case studies, technological breakthroughs, and industry trends as a guide to understanding these impediments to cloud adoption. The study reviews the theoretical background of cloud computing, which include models related to deployment and service, and illustrates its successes in performance, reliability, and security. Persistent barriers, however, mean security and compliance are insecure while cost unpredictability remains a concern for organizations in using the maximum potential of a cloud. The study proposes strategic solutions that can serve to bridge the gap in terms of hybrid and multi-cloud adoption, AI-driven security frameworks, regulatory compliance automation, and various cost optimization techniques. Emerging trends like quantum computing, edge computing, and green cloud initiatives are shaping the future of cloud computing. By implementing these solutions, more is made out of the potential of cloud computing in securing a more efficient and sustainable digital infrastructure. 2025 IEEE. -
Mitigating Subjectivity and Annotation Inconsistencies in Sentiment Analysis via an SVM-RoBERTa Ensemble
This research addresses a main limitation in the Natural Language Processing that is the impact of subjectivity and annotation inconsistencies on the accuracy of the sentiment classification. We did a systematic comparison of two fundamentally different architectures. A traditional feature based Support Vector Machine and a deep contextual fine tuned RoBERTa transformer using a challenging, noisy, real-world Twitter dataset. This corpus retains ambiguity and sarcasm on purpose and serve the crucible for testing model robustness. We developed a soft voting ensemble method that combines the probability scores from both models to obtain the best predictive capabilities. The results showed a clear technological hierarchy. The RoBERTa model with its deep semantic grasp outperformed the SVM by a substantial margin achieving 90% accuracy against 83.5% accuracy. But the hybrid ensemble model attained the highest overall accuracy of 91.35% and showed better reliability across all the sentiment classes. These findings shows that a hybrid approach fusing a transformer's nuanced understanding with the stabilization provided by ensemble learning is the most effective and robust method for mitigating data imperfections in modern sentiment analysis. 2025 IEEE. -
Hierarchical Mapping-Partitioning-Search with Attention-Weighted Communication for UAV Swarms in Search and Rescue Operations
UAV swarm Search and Rescue (SAR) operations demand intelligent coordination to function efficiently in unfamiliar terrains while maintaining communication under bandwidth limitations. To address this, we propose a Hierarchical Mapping-Partitioning-Search (HMPS) framework that combines quadtree-based adaptive partitioning of the search area with deep reinforcement learning for region selection, together with an Attention-Weighted Flooding (AWF) communication protocol to enhance swarm coordination. The HMPS framework adapts search granularity to uncertainty and obstacle density, uses a Deep Q-Network (DQN) to learn a region-selection policy, and employs a lightweight local coverage planner to improve exploration efficiency. The AWF protocol prioritizes message relays based on content and link quality, reducing bandwidth while preserving essential information flow. This paper presents HMPS as a practical option for autonomous swarm SAR operations, and reports encouraging preliminary results in GPS-denied terrains. 2025 IEEE. -
Exploring Communication Authenticity Anxiety: A Data-DrivenPsychological Analysis of Al-Generated Content on StudentSelf-Perception and Expression
Generative artificial intelligence (AI) tools such as ChatGPT and Gemini are becoming more common in student communication, owing to the improvement that they offer in fluency and efficiency, but at the same time raise concerns about authenticity. Students struggle to put their authentic voice forward in the quest to enhance their work using these writing assistants. Many surveys have been conducted, which indicate widespread use of AI tools for education-related chores, yet these studies ignore the emotional effects related to this. The psychological discomfort related to authenticity in text-based communication is still not well examined and to address this gap, this study introduces a term called Communication Authenticity Anxiety and successfully examines its relationship with self-perception, academic stress, resilience, and AI dependence. Data were collected via a structured student survey and analyzed using exploratory factor analysis, regression modeling and machine learning techniques. Results show that self-perception and academic stress are the strongest predictors of authenticity anxiety, while resilience and AI dependence have weaker effects. These findings were further validated by Machine Learning models, with Random Forest achieving 75% accuracy and XGBoost achieving {9 2%}. This study, thus, successfully contributes to understanding the various psychological consequences of AI-generated content on student identity and expression, thereby providing valuable insights for crafting responsible educational policies. 2025 IEEE. -
Exploring Communication Authenticity Anxiety: A Data-DrivenPsychological Analysis of Al-Generated Content on StudentSelf-Perception and Expression
Generative artificial intelligence (AI) tools such as ChatGPT and Gemini are becoming more common in student communication, owing to the improvement that they offer in fluency and efficiency, but at the same time raise concerns about authenticity. Students struggle to put their authentic voice forward in the quest to enhance their work using these writing assistants. Many surveys have been conducted, which indicate widespread use of AI tools for education-related chores, yet these studies ignore the emotional effects related to this. The psychological discomfort related to authenticity in text-based communication is still not well examined and to address this gap, this study introduces a term called Communication Authenticity Anxiety and successfully examines its relationship with self-perception, academic stress, resilience, and AI dependence. Data were collected via a structured student survey and analyzed using exploratory factor analysis, regression modeling and machine learning techniques. Results show that self-perception and academic stress are the strongest predictors of authenticity anxiety, while resilience and AI dependence have weaker effects. These findings were further validated by Machine Learning models, with Random Forest achieving 75% accuracy and XGBoost achieving {9 2%}. This study, thus, successfully contributes to understanding the various psychological consequences of AI-generated content on student identity and expression, thereby providing valuable insights for crafting responsible educational policies. 2025 IEEE. -
An Enhanced A3C-LSTM Framework with Attention for Dynamic Portfolio Allocation in Equity Markets
Portfolio optimization in dynamic financial markets presents a significant challenge for traditional models. This paper introduces an advanced deep reinforcement learning framework for portfolio management based on an enhanced Asynchronous Advantage Actor-Critic (A3C) algorithm. This paper integrate's a Long Short-Term Memory layer and a multi-head attention mechanism into the actor-critic architecture to more effectively capture temporal dependencies and feature importance within financial time-series data. The model's novelty lies in its enriched state representation, which includes a comprehensive set of technical indicators, inter-asset correlation matrices, and market regime analysis. Furthermore, we employ a sophisticated riskadjusted reward function, incorporating penalties for drawdown and volatility alongside a bonus based on the Sortino ratio. The agent was trained and tested in a simulated environment using historical daily price data from five major S&P 500 stocks. Experimental results demonstrate that our agent successfully learns a robust and adaptive allocation strategy, significantly outperforming an equal-weight benchmark in terms of overall return, Sharpe ratio, and maximum drawdown. This study underscores the potential of sophisticated DRL architectures to navigate complex market dynamics and optimize for riskadjusted performance. 2025 IEEE.