Browse Items (3095 total)
Sort by:
-
Breast Cancer Diagnosis: Feature Selection and Ensemble Machine Learning
Breast cancer diagnosis requires accurate diagnostic tools that are both efficient and interpretable for clinical deployment. This study presents an integrated pipeline combining Recursive Feature Elimination with Cross-Validation (RFECV), Synthetic Minority Over-sampling Technique (SMOTE), and ensemble learning methods applied to the Wisconsin Breast Cancer Diagnostic dataset. RFECV achieved dimensionality reduction from 30 to 17 features, representing a 43% reduction while maintaining predictive performance. SMOTE transformed the class imbalance ratio from 1.68:1 to a perfect 1:1 balance. A comprehensive evaluation of twelve machine learning models revealed that LightGBM attained an F1-score of 0.9722, accuracy of 96.5%, and ROC-AUC of 0.9914 with strong cross-validation stability (0.9681 0.0179). Feature importance analysis identified worst perimeter, area, and concave points as the most discriminative features for differentiating malignant from benign tumors. The proposed approach achieved a 35% reduction in training time compared to full-featured models without sacrificing performance. This reproducible pipeline demonstrates practical clinical relevance for automated breast cancer diagnosis with improved computational efficiency and model interpretability. 2025 IEEE. -
Fine-Tuning Large Language Models for Personality Development
Large Language Models (LLMs) are state-of-the-art in Natural Language Processing (NLP) tasks but are extremely challenging to fine-tune with their size and computational needs. The current research targets the fine-tuning of the OpenLLaMA-3B model for personality development tasks with parameter-efficient fine-tuning (PEFT) via LoRA and QLoRA. To deal with hardware limitations, 4-bit quantization through BitsAndBytes was used, lowering memory without affecting accuracy. A FAISS-based Retrieval-Augmented Generation (RAG) pipeline was also used to improve contextual reliability. Semantic similarity (cosine similarity), BLEURT, ROUGE, and human evaluation were used to evaluate the model. The results of this experiment show that semantic similarity greatly increasing from 0% before fine-tuning to 80% after fine-tuning, demonstrating the benefit of PEFT and quantization in domain adaptation. The work illustrates that by leveraging effective fine-tuning, quantization, and retrieval augmentation, LLMs can be deployed at scale under limited resources while providing high contextual accuracy for personality development purposes. This approach not only enhances knowledge of the job, but it also has practical scalability for educational and self-improvement purposes. The findings highlight a viable path forward for applying small, yet strong, LLMs to personalized learning and adaptive human-AI interaction. 2025 IEEE. -
A Hybrid Ensemble Model Combining Machine Learning and Fuzzy Logic for Robust Stress Level Detection
Accurate stress detection remains challenging because stress manifests differently across diverse populations. This research presents a novel hybrid approach that combines four machine learning classifiers - XGBoost, Random Forest, Support Vector Machine, and Logistic Regression - with an enhanced fuzzy logic system through weighted voting. The study employed a comprehensive dataset comprising over 1,100 samples containing more than 45 features that capture psychological, academic, social, physical, and environmental stress indicators. The results demonstrate that the ensemble achieves 87.27% accuracy with statistically significant improvements over individual models (p0.001). The approach maintains strong performance when 10% noise is injected or missing data is simulated, and provides well-calibrated probability estimates that could support clinical decision-making. 2025 IEEE. -
High-Efficiency Five-Level Multilevel Inverter for Grid-Connected Renewable Energy Systems: Modeling and Simulation
This paper describes the design of a 5-level multilevel inverter for grid connected renewable energy systems simulated using MATLAB. The inverter developed works with an input voltage of 164 V and a well-regulated output voltage of 325.4 V, with the switching frequency fixed at 20 kHz for efficiency, while controlling the modulation index (M) for proper voltage regulation. The performance review of the inverter such as THD and power factor were simulated with MATLAB. It is found that the voltage quality has been greatly improved and harmonic vicious generation has been diminished when compared with traditional inverter. The high-quality of output voltage provided by the inverter also makes it ideal for use with renewable energy systems. The simulations also demonstrate the efficiency of the inverter when controlling power for grid connection, revealing its capability to offer secure and low-cost renewable energy solutions. Performance evaluations indicate improved voltage waveform quality and higher efficiency of the system, which makes it promising for future renewable energy connected applications. The fact of using MATLAB simulations to verify our designs further proves its feasibility and efficacy in handling the challenges of power conversion in sustainable energy systems. 2026 IEEE. -
Intelligent YOLOv8-based Rover for Precision Agriculture: Tomato Ripeness and Disease Detection
Tomato cultivation has traditionally relied on manual inspection and generalized irrigation practices, often resulting in inefficient resource use and reduced yield quality. This work presents a compact rover designed to support precision agriculture in tomato farming. The rover is equipped with an OV2640 camera module, a humidity sensor, and a water-level sensor. The OV2640 captures images of tomato fruits and leaves, transmitting them via Wi-Fi to a laptop for analysis. Two custom-trained YOLOv8 deep learning models are employed for visual diagnostics: one determines tomato ripeness, enabling optimal harvest timing, and the other detects common leaf diseases, including Late Blight, Leaf Mold, Leaf Miner, Mosaic Virus, Septoria, Early Blight, Spider Mites, and Yellow Leaf Curl Virus. In addition to visual inspection, the rover measures environmental parameters such as soil moisture and ambient humidity, supporting data-driven irrigation decisions and early preventive measures. Communication between the rover and the processing unit is achieved through live video streaming from the ESP32-CAM, with processed results enabling either manual teleoperation or potential future autonomous navigation. By integrating AI-based plant health assessment with environmental monitoring, the proposed system offers a economically efficient and scalable solution to improve crop quality, optimize resource usage, and enhance decision-making in tomato farming operations. 2026 IEEE. -
Reinforcement Learning-Driven Energy Management for Battery-Supercapacitor Hybrid Storage in Electric Vehicles
The fast growth of the electric vehicles (EVs) market has increased the requirements towards high power transients, efficiency, and reliability on automotive onboard energy management systems by extending battery lifetime. Pure battery storage systems are similarly subject to frequent peak power demands during rapid acceleration and regenerative braking, and thus suffer from rapid aging. Aiming at this issue, in this paper, an AI-based EMS for a battery-supercapacitor HESS in EVs is developed. Dynamic driving conditions are handled by an RL-based power splitting control strategy which dynamically divides power between lithium-ion battery and supercapacitor in this context. The battery stress is to be minimized with the stabilization of the DC-link voltage and traction power demand. System modeling and validation is carried out in MATLAB/Simulink with the use of typical urban drive cycles. Simulation results show that, compared with a rule-based control of the EMS, our proposed AI-enabled EMS can decrease battery peak current by 38.6%, enhance energy efficiency by 11.2%, and increase cycle life by around 27%. The deviation of the DC-link voltage is limited within 1.8% and such control can be used to reduce total system response time in rapid load transition by 22%. Comparison results reveal that the optimal management framework has better adaptability and stability when compared to the corresponding one under different loads and driving conditions, which are promising for next generation EVs energy management issues. 2026 IEEE. -
DC-DC and DC-AC Converters with Bi-Directional Capabilities for EV Applications
Abstract: This paper presents the design and implementation of a compact bidirectional DC-DC converter coupled with a DC-AC inverter for electric vehicle (EV) motor-drive applications. Both propulsion and regenerative braking modes are made possible by the suggested architecture, which facilitates smooth power transfer between a 48 V battery and a series- wound AC motor. Additionally, the inverter provides controlled AC stimulation for a dependable motor operation, while a high- efficiency bidirectional DC-DC converter controls battery power flow during acceleration and recovers energy during braking. In order to maintain dynamic stability under changing load circumstances, an Arduino Nano microcontroller uses a proportional-integral (PI) control method to regulate motor speed and current. Cross-conduction losses are decreased and MOSFET switching safety is improved by customized PWM pulse generation with dead-time insertion. 2026 IEEE. -
Blockchain-Driven Architecture for Decentralized Energy Transactions in Smart Grids
The versatility of blockchain technology enables its capabilities to protect decentralized energy trading and transform modern smart grids by removing all dependencies on centralized utility operators and eliminating vulnerabilities that stem from data tampering, pricing manipulation, and single-point failures. The purpose of this paper is to present a fully virtualized and software-implemented architecture of a blockchain that incorporates a lightweight Proof-of-Authority (PoA) consensus model, dynamic pricing smart contract(s), and a multi-layer energy ledger, tailored specifically for seamless peer-to-peer energy trading. The proposed energy trading model is built using an entirely virtualized architecture and is validated through simulation, as opposed to previously proposed models that are based on expensive consensus mechanisms and require hardware-assisted metering. The proposed model delivers significant improvements (37.4% reduction in transaction latency, 52.8% improved throughput, and 41.6% lower computational overhead) when compared to traditional Proof-of-Work and DAG-based models. The smart contract engine ensures energy-pricing fluctuations remain stable, and the system as a whole achieves 95.2% transaction validity, all while preserving ledger immutability, user anonymity, and high scale performance. The results achieved from this innovative software-defined architecture ensure its decentralized smart-grid deployments and high scalability exceed market expectations. 2026 IEEE. -
Exploring Conditional Generative Models for Sketch-to-Image Translation: cGAN, cVAE, and Conditional Diffusion Models
Creating realistic facial pictures from hand-drawn sketches is of significant utility in forensic investigations because eyewitness drawings are frequently the only visual leads for suspect identification. Turning a hand-drawn sketch into a realistic image is a difficult task. This is because sketches lack detailed information, they are abstracted, and ambiguous. Most of the conventional image creation and generation techniques tend to lose facial structure, identity, and realism. This makes it a great area for generative AI. This paper is a comparative analysis of three generative models: Conditional GANs, Conditional VAEs, and Conditional Diffusion Models. We evaluate these models on the sketch-to-image synthesis problem using the CUHK Face Sketch Dataset. We recognize and compare how every model handles the challenge of generating images from sketches of faces, with an emphasis on producing realistic images, maintaining identity and diversity. The paper demonstrates the advantages and disadvantages of each approach. It also offers insights into their usefulness for forensic applications and suggests directions for future improvements through combined or specialized generative structures. 2025 IEEE. -
Exploring Conditional Generative Models for Sketch-to-Image Translation: cGAN, cVAE, and Conditional Diffusion Models
Creating realistic facial pictures from hand-drawn sketches is of significant utility in forensic investigations because eyewitness drawings are frequently the only visual leads for suspect identification. Turning a hand-drawn sketch into a realistic image is a difficult task. This is because sketches lack detailed information, they are abstracted, and ambiguous. Most of the conventional image creation and generation techniques tend to lose facial structure, identity, and realism. This makes it a great area for generative AI. This paper is a comparative analysis of three generative models: Conditional GANs, Conditional VAEs, and Conditional Diffusion Models. We evaluate these models on the sketch-to-image synthesis problem using the CUHK Face Sketch Dataset. We recognize and compare how every model handles the challenge of generating images from sketches of faces, with an emphasis on producing realistic images, maintaining identity and diversity. The paper demonstrates the advantages and disadvantages of each approach. It also offers insights into their usefulness for forensic applications and suggests directions for future improvements through combined or specialized generative structures. 2025 IEEE. -
Pattern Reconfigurable Antennas for Wireless Applications: A Review of Design Techniques and Advances
This literature survey investigates advances in pattern reconfigurable antennas that take advantage of Meta-surface (MS) technology over other techniques. The study explores how these reconfigurable antennas are transforming next-generation communication systems, addressing critical applications such as 5G/6G networks and smart wireless environments. Key research observations include the ability of MS-based designs to achieve dynamic beam steering, crucial to meeting the diverse requirements of future communication systems. The paper also identifies challenges such as design complexity, power efficiency, integration with existing systems, and scalability for practical deployments. By highlighting these advances and addressing open challenges, this survey aims to provide information on the potential of MS-enabled reconfigurable antennas to shape the future of wireless communication technologies. 2025 IEEE. -
Using Machine Learning Sentiment Analysis to Evaluate Students Learning Impact
For educational experiences and results to be improved, learning impact assessment is essential. Students' emotional reactions, which are crucial to their involvement and understanding, are frequently missed by traditional evaluation techniques. Through a review of student feedback, conversations, and course ratings, this study investigates the use of machine learning-based sentiment analysis to assess the impact of learning. Performance evaluations were conducted on a number of sentiment categorization models, including Nae Bayes, Support Vector Machines (SVM), Logistic Regression, Random Forest, Long Short-Term Memory (LSTM), and BERT. With an accuracy of 91.7%, the results show that BERT performs better than other models and offers more accurate sentiment classification. Accuracy and insights are further improved by combining textual, auditory, and visual signals in multi-modal sentiment analysis. The results show how sentiment analysis may be used to track feedback in real time facilitating adaptive learning techniques to raise student interest. Future studies should concentrate on expanding sentiment analysis applications to traditional and hybrid learning contexts, integrating multi-modal data, and ethical implications. 2025 IEEE. -
A Systematic Approach for Predicting Cybersecurity Attacks in IoT using CNN-LSTM with HABCABO
IoT has transformed how devices work together. Now, billions of connected devices may share data across smart homes, energy systems, and environmental monitoring. In Internet of Things ecosystems, rapid IoT expansion has made them very vulnerable, which makes them easy targets for cyberattacks. Hackers can break into IoT devices that don't have enough protection to stop services, steal data, and invade privacy. This paper shows how to use deep learning using CNNs and LSTM networks and the HABCABO optimization algorithm to deal with these new dangers. After careful sequencing, scaling, and noise reduction, filter-based feature selection uses statistical methods to keep the most important information. To get the best detection, the CNN-LSTM model is trained with features that are carefully regulated. The suggested model is more accurate than CNN and LSTM approaches, with an accuracy rate of 98.04 %. These results show that the model can find and stop IoT cybersecurity threats. In conclusion, CNN-LSTM and HABCABO are strong and smart ways to make sure that IoT infrastructure is safe and reliable right now. 2025 IEEE. -
Smart Facial Expression Analysis: Fuzzy Extreme Learning Machine in Emotion Detection
The immense academic and economic potential of facial emotion recognition (FER) has made it a crucial field in computer vision and artificial intelligence. Because of the fundamental role that facial expressions play in interpersonal communication; face photographs are vital for analysing human emotions within the context of Smart Facial Expression Analysis. This research provides a successful pipeline for emotion identification and examines FER methods that rely just on face pictures. Preprocessing, segmentation, feature extraction, and training the model are the steps that make up the suggested method's organised procedure. Face detection using the Viola-Jones technique is the first step in the preprocessing phase. Four rectangular characteristics are used for segmentation, with greyscale conversion being a necessity. In order to train a fuzzy -ELM model, feature extraction uses HOS and LBP. Emotions are better understood with this method. The suggested fuzzy-ELM approach outperforms two state-of-the-art models, ELM and CNN. With an accuracy of 98.33 %, the experimental findings show a substantial improvement in precision. A dependable and high-performing method for emotion recognition using just facial imaging, these findings highlight the usefulness of the suggested approach for Smart Facial Expression Analysis. 2025 IEEE. -
A Novel Network-Based Digital Payment Fraud Detection using OP-ELM Network
Internet and Industry 4.0 have helped banks and other financial organizations enhance procedures and decrease fraud. Digital payment techniques have helped internet buying skyrocket. Industry 4.0 promotes process optimization, ecosystem collaboration, and growth by integrating digital systems with physical and IoT devices. Unfortunately, digital payment cybercrime has grown rapidly, causing large annual financial losses. Because of this, fraud detection systems must be constantly improved. The suggested TLELM approach includes preprocessing, feature selection, and model training. Preprocessing involves standardizing data, eliminating outliers, and handling null or missing values. The CSO technique selects relevant features by optimizing selection. A new approach combining TL and ELM improves DPFD procedures. The new metaheuristic TL excels at combinatorial optimization. TLELM efficacy was examined using multiple datasets. The recommended method was compared to top-tier algorithms for binary and multiclass data categorization. Experimental data shows that TLELM outperforms other models with 99.37% accuracy. This study found that TLELM can detect online payment fraud. The method optimizes fraud detection and classification accuracy using TL and ELM. Add more real-world datasets to strengthen robustness and make additional improvements to handle future fraud methods. 2025 IEEE. -
Deploying a Multi-Model Forecasting System for Bitcoin Prices: Bridging Statistical Forecasting and Deep Learning Innovations
In this study, we investigate and compare several forecasting models for predicting Bitcoin market prices using historical data sourced from Nasdaq Data Link (formerly Quandl) spanning from 2016 to 2025. Our analysis evaluates traditional time series methods - such as ARIMA and Holt Winters exponential smoothing - alongside modern machine learning and deep learning techniques including LSTM, Prophet, XGBoost, SVR, Random Forest, and GRU. Performance was assessed via metrics such as RMSE, MAE, MAPE, sMAPE, directional accuracy, and R-squared. Our experiments reveal that while classical methods (e.g., ARIMA and Holt Winters) exhibit large estimation errors and limited explanatory capacity, advanced neural network architectures - particularly the GRU - demonstrate superior accuracy with an RMSE of 2,505.84, MAE of 1,760.93, MAPE of 2.79%, and an R-squared of 0.99. The best-performing model (GRU) was deployed as a web application on PythonAnywhere, providing real-time forecasts through an interactive dashboard. This deployment not only validates the predictive efficacy of the GRU model but also offers a practical tool for investors and financial analysts to monitor and predict Bitcoin price movements using reliable Nasdaq data. 2025 IEEE. -
Image Data Driven Lung Cancer Classification using Deep Learning and Optimization Algorithm
Lung cancer continues to be a major contributor to global cancer mortality, underscoring the importance of early detection and accurate diagnosis. This work introduces an integrated framework that leverages deep learning in combination with Bayesian optimization to achieve robust lung cancer classification. Convolutional Neural Networks (CNNs) are employed for feature extraction and image analysis, while Bayesian optimization is applied to automatically fine-tune critical hyperparameters, thereby improving accuracy and minimizing training overhead. The methodology focuses on the analysis of computed tomography (CT) images to distinguish between different lung cancer categories. By addressing the limitations of manual hyperparameter selection, the proposed framework enhances the efficiency and reliability of deep learning models in medical imaging. The outcomes of this study highlight its potential contribution to computer-aided diagnosis, offering clinicians an effective decision-support tool for precise and timely lung cancer detection. 2025 IEEE. -
Comparative Analysis of Machine Learning Algorithms for Effective Crop Recommendation
The global call for sustainable farming necessitates a move away from traditional crop selection methods. These conventional approaches, often relying on farmer intuition, are imprecise and scale poorly in the face of complex environmental variables. Machine Learning (ML) models offer a robust, data-driven solution. By analyzing multifaceted data-spanning soil chemistry, weather patterns, precipitation trends, and historical yield performance-ML models can significantly enhance decision-making, optimize resource utilization, and improve overall crop outcomes. This paper delivers an extensive comparative review of key ML algorithms employed for crop recommendation, including Decision Trees (DT), Random Forests (RF), Support Vector Machines (SVM), k-Nearest Neighbors (KNN), and Artificial Neural Networks (ANN). We also explore the critical role of Explainable AI (XAI) in building model transparency. Our study evaluates these models on the metrics of accuracy, interpretability, and computational overhead. The research also investigates hybrid methods that integrate deep learning with conventional ML to enhance predictive power. Our comparative findings highlight the strengths and weaknesses of each model, concluding that ANN and XAI-based approaches demonstrate the highest accuracy and adaptability for diverse agricultural conditions. We also identify significant challenges, such as data imbalances and the absence of real-time data, and discuss future trends like the integration of IoT, remote sensing, and federated learning, which will be key to making precision farming scalable and accessible. 2025 IEEE. -
AI-Driven Continuous Learning Analysis and Blockchain Validation: A Review on Innovations in Digital Education
The rapid evolution of digital education has necessitated innovative approaches to enhance learning experiences, provide personalized insights, and ensure the credibility of achievements. This study analyses the challenges in AI-based continuous learning analysis for students and how we can securely share certificates through blockchain technology. The amalgamation of artificial intelligence and blockchains can create a secure, open, and trustworthy environment that schools, students, and employers can make use of. It is important to protect student data along with technological advancements. In response to the dynamic landscape of digital education, novel approaches are essential to enrich learning experiences, offer personalized insights, and maintain the credibility of achievements. This research reviews existing AI-based framework that facilitates ongoing learning assessment for students while ensuring secure certificate sharing through blockchain technology. By amalgamating artificial intelligence and blockchain, a robust ecosystem emerges that fosters transparent, efficient, and secure interactions among educational institutions, learners, and employers. Drawing from these evaluations, the framework suggests customized learning paths, thus amplifying the effectiveness of learning journeys. The objective of this study is to analyze the existing methods and to suggest the latest technologies in education, which will help the education sector to keep track of technological advancements and be in the race. 2025 IEEE. -
Prediction of Facial Emotions using Deep Learning and Machine Learning Techniques
Facial expression prediction has gained considerable attention in recent years, particularly because of its applications in human-computer interaction. This paper compares a wide range of deep learning and machine learning models for the prediction of face emotion using the CK+ dataset proposed by Cohn-Kanade. The dataset is characterized by seven classes of emotions represented by the labels, namely surprise, happiness, disgust, anger, sadness, fear, and contempt, on 784 training, 98 validation, and 99 testing images. To further improve model performance, preprocessing techniques were employed that enhanced data efficiency. To increase variability in the data and reduce overfitting, all images were scaled to a 48*48 pixel resolution, pixel values were scaled to be between 0 and 1 for uniformity and the following data augmentation techniques were implemented: 10-degree rotation, horizontal flip, 0.15 zoom. The five models that were tested were CNN, SVM, VGG16, InceptionV3 and VGG19. The results demonstrate the high accuracy achieved by the CNN model which showed an accuracy of 98.98%, 99% and 99% in training, validation and test respectively. The SVM classifier got an accuracy of 99%. Both InceptionV3 and VGG19 on the other hand achieved competitive testing accuracy values of 90.91% and 97.98% respectively, while VGG16 got tested and reached an accuracy of 85.86%. 2025 IEEE.