Browse Items (16481 total)
Sort by:
-
A Hybrid Clustering Approach for Enhanced Classification Efficiency in Data Analytics
Clustering is a fundamental technique in data analytics that groups data points with similar characteristics into clusters. It is crucial for uncovering hidden patterns, trends, and structures in datasets. Clustering reduces the complexity of large datasets by summarizing data into representative clusters. This simplification makes it easier to analyze and interpret data, especially when dealing with high-dimensional datasets. By identifying meaningful groups, clustering provides actionable insights that supports decision-making. For instance, businesses can make concrete decisions about product recommendations, pricing strategies, or resource allocation based on cluster analysis. The approach described in the paper offers an efficient method for combining K-means and Gaussian Mixture Model (GMM) clustering techniques. The method combines two wellknown clustering techniques, K-means and GMM, to leverage their respective strengths. K-means is known for its simplicity and efficiency, while GMM can model complex data distributions with varying covariance structures. Instead of directly integrating the results of K-means and GMM, the approach uses a simplified averaging technique to converge the cluster labels obtained independently from both methods. This suggests that the method may involve assigning weights to the cluster labels obtained from K-means and GMM and then averaging them to obtain final cluster assignments. Overall, this approach presents a promising direction for combining K-means and GMM clustering techniques, offering a streamlined integration process that simplifies the consideration of varying covariance types in GMM. The effectiveness of the method is evaluated through empirical studies and comparisons with existing clustering approaches. 2025 IEEE. -
A hybrid crypto-compression model for secure brain mri image transmission
Medical image encryption is a major issue in healthcare applications where memory, energy, and computational resources are constrained. The modern technological architecture of digital healthcare systems is, in fact, insufficient to handle both the current and future requirements for data. Security has been raised to the highest priority. By meeting these conditions, the hybrid crypto-compression technique introduced in this study can be used for securing the transfer of healthcare images. The approach consists of two components. In order to construct a cutting-edge generative lossy compression system, we first combine generative adversarial networks (GANs) with oearned compression. As a result, the second phase might address this problem by using highly effective picture cryptography techniques. A randomly generated public key is subjected to the DNA technique. In this application, pseudo-random bits are produced by using a logistic chaotic map algorithm. During the substitution process, an additional layer of security is provided to boost the techniques fault resilience. Our proposed system and security investigations show that the method provides trustworthy and long-lasting encryption and several multidimensional aspects that have been discovered in various public health and healthcare issues. As a result, the recommended hybrid crypto-compression technique may significantly reduce a photos size and remain safe enough to be used for medical image encryption. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023. -
A Hybrid Deep Learning and Ensemble Framework for Real-Time Cyclone Path and Intensity Prediction in Disaster-Prone Regions
Predicting the path and strength of cyclones involves significant issues in meteorology, as mistakes can greatly affect disaster management and evacuation strategies. Current models frequently encounter difficulties in achieving accurate real-time forecasting, particularly in representing complicated spatial-temporal dynamics of cyclones. The proposed study presents an innovative hybrid architecture combining deep learning and ensemble methods, using convolutional layers, LSTM units, and a gradient boosting meta-learner to improve prediction efficiency. The system was trained and verified utilising multi-year cyclone datasets obtained from Kaggle, which included atmospheric and oceanic factors. The model architecture attained exceptional accuracy, with a track error of 28 km, a mean absolute error (MAE) of 3.2 hPa for pressure, 4.5 km/h for wind speed, and a root mean square error (RMSE) of 35.4 km. The suggested approach consistently outperformed baseline models, including ConvLSTM, GRU, and XGBoost, across all critical criteria. The deployment in real-time was enabled by a containerised, low-latency API that can integrate with disaster early warning systems. This research enhances cyclone forecasting by offering a scalable, precise, and operationally feasible solution for disaster-prone areas, demonstrating practical superiority over current methodologies. The results highlight the capability of hybrid AI models to improve the accuracy and dependability of meteorological forecasts. 2025 IEEE. -
A hybrid deep learning and quantum computing approach for optimized encryption algorithms in secure communications
As online dangers get worse, there is a greater need for strong encryption methods to protect private conversations. Utilizing the strengths of both deep learning and quantum computing, this study suggests a new mixed method for improving the security of communication systems by making encryption algorithms work better. When it comes to keeping up with new online threats, traditional security methods often fall behind. Deep learning techniques could be a good way to improve encryption algorithms because they let the system learn and change to new attack methods. In the meantime, quantum computing offers unmatched computing power that can completely change how cryptography works by using quantum events like superposition and entanglement. Our suggested method combines the flexibility of deep learning with the computing power of quantum computing to get around the problems with current encryption methods. This will make safe communication systems more resistant to attacks from smart people. Through tests and models, we show that our mixed approach works better and more effectively than current encryption methods. This shows that it has the ability to solve the growing safety problems in a world that is becoming more and more linked. 2024, Taru Publications. All rights reserved. -
A Hybrid Deep Learning Model Using U-Net and Vision Transformer for Artificial Intelligence Powered Cervical Stenosis Diagnosis
This study presents a deep learning-based approach for the classification of cervical stenosis using MRI spine images, integrating multiple phases such as preprocessing, segmentation, feature extraction, and classification. A U-Net-based segmentation model effectively delineates key anatomical structures, including the spinal canal, intervertebral discs (IVDs), and neural foramen, improving feature extraction and classification accuracy. Furthermore, ResNet-50 is employed for feature map generation, leveraging deep hierarchical representations to extract meaningful spatial patterns from MRI slices. For classification, a Vision Transformer (ViT)-based model is utilized, taking advantage of its self-attention mechanism to capture both local and global dependencies within MRI images. Unlike conventional CNN-based models, ViT processes MRI scans as patches, enabling a more context-aware analysis of stenotic regions. The model is trained using an 80%20% train-test split and evaluated using standard performance metrics, achieving an accuracy of 92.60%, precision of 90.16%, recall of 95.43%, and an F1-score of 91.56%. These results indicate that the ViT model outperforms traditional CNN-based classifiers in cervical stenosis detection, ensuring higher sensitivity and specificity in real-world clinical applications. The Author(s), under exclusive license to Springer Nature Switzerland AG 2026. -
A Hybrid Deep-ensemble Decision-Support Framework for Reliable Early Breast Cancer Detection: A Cross-validated Outcome Analysis
OBJECTIVE The necessity to diagnose breast cancer early and correctly is the need to minimize the diagnostic uncertainty and unwarranted clinical procedures. This paper assesses the reliability of a hybrid deep-ensemble decision-support model in terms of diagnostic reliability, stability of outcome, and translational feasibility of the model via structured clinical data to detect early breast cancer. METHODS The Wisconsin Diagnostic Breast Cancer dataset which consisted of 569 cases of benign and malignant tumors was analyzed retrospectively. The framework proposed combines the deep learning of latent representations with stacked classification, ensemble-based feature selection, and stacked classification. Performance evaluation was performed based on sensitivity, specificity, accuracy, F1-score, and area under the curve (AUC) performed using stratified 10-fold cross-validation. The statistical stability across folds and the comparison with baseline models were determined with the help of non-parametric tests (p<0.05). RESULTS The model had good diagnostic performance with an accuracy of between 91.2-100 (Mean 96), Sensitivity of 76.2-100, good specificity value, and AUC 0.973-1.000. Variability in performance between folds was low, and statistically significant enhancement as compared to baseline classifiers were present. CONCLUSION The hybrid deep-ensemble model is highly diagnostic, has robust discriminative ability, and ultimately remains stable, which demonstrates the methodological robustness and diagnostic reliability of the proposed framework as a proof-of-concept decision-support model for early breast cancer detection, with potential translational relevance subject to further external clinical validation. 2026, Turkish Society for Radiation Oncology. -
A Hybrid Edge-Cloud Computing Approach for Energy-Efficient Surveillance Using Deep Reinforcement Learning
This paper explores the novel application of Deep Reinforcement Learning (DRL) in designing a more efficient, scalable, and distributed surveillance architecture, which addresses concerns such as data storage limitations, latency, event detection accuracy, and significant energy consumption in cloud data centers. The proposed architecture employs edge and cloud computing to optimize video data processing and energy usage. The study further investigates the energy consumption patterns of such a system in detail. The implementation leverages machine learning models to identify optimal policies based on system interactions. The proposed solution is tested over an extensive period, resulting in a system capable of reducing latency, enhancing event detection accuracy, and minimizing energy consumption. 2023 IEEE. -
A hybrid ensemble framework with particle swarm optimization for network anomaly detection
The increasing complexity of cyber threats necessitates the development of a robust and adaptive Intrusion Detection System (IDS) capable of safeguarding network infrastructures. Traditional IDS approaches often struggle to detect sophisticated attacks due to their reliance on predefined patterns. We propose an adaptive particle swarm optimization (PSO)-optimized ensemble learning framework tailored to address these challenges in modern IDS applications. Our approach leverages the NSL-KDD and CICIDS datasets to ensure the IDS is trained and evaluated on data reflecting current network behaviours and threat landscapes. We evaluate multiple machine learning models, including Decision Trees (DT), Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), Random Forests (RF), and an ensemble of these models for both binary and multi-class classification tasks. By incorporating adaptive mechanisms within the PSO algorithm, our framework dynamically adjusts hyperparameters during optimization, enhancing model robustness and convergence speed. The proposed framework is also benchmarked against state-of-the-art IDS approaches, including ASRL and PSOGSA. Empirical evaluations demonstrate that the ensemble model achieves superior detection accuracy and reduced false positive rates, thereby advancing the efficacy of intrusion detection methodologies. The Author(s) 2025. -
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. -
A Hybrid Evolutionary Deep Learning Model Integrating Multi-Modal Data for Optimizing Ovarian Cancer Diagnosis
This research intends to enhance ovarian cancer detection by the combination of state-of-the-art machine learning algorithms with extensive multi-modal datasets. The Convolutional Neural Network (CNN), K-Nearest Neighbors (KNN), and VGG16 models were thoroughly assessed, displaying remarkable precision, recall, F1 scores, and overall accuracy. Notably, VGG16 emerged as a strong performance with a precision of 0.97, recall of 0.96, F1 score of 0.97, and accuracy reaching 98.65%. The addition of confusion matrices enables a thorough insight on each model's classification performance. Leveraging multiple datasets, spanning CT and MRI scans with demographic and biographical facts, promotes the holistic knowledge of ovarian cancer features. While the suggested Hybrid Evolutionary Deep Learning Model was not deployed in this work, the results underscore the potential for its development in future research. These discoveries signify a huge leap forward in early detection capabilities and individualized treatment techniques for ovarian cancer patients. As technology and medicine combine, this study tracks a road for breakthrough diagnostic approaches, empowering clinicians and encouraging favourable results in the continuing struggle against ovarian cancer. 2024 IEEE. -
A Hybrid FinTech Fraud Detection Model Integrating Multiscale Entropy and Transformer-GAT Techniques
Credit card fraud has become a major concern in the FinTech industry due to the rapid growth of digital payment platforms and the increasing sophistication of fraudulent activities. Accurate and timely detection of fraud is essential to minimize financial losses and maintain trust in FinTech services. This study presents a hybrid deep learning framework for credit card fraud detection using the 2023 Credit Card Fraud Detection Dataset. The proposed approach with data preprocessing, which includes handling missing values, removing duplicate entries, and encoding categorical features to ensure clean and structured input for modeling. Normalization is applied to scale features uniformly, preventing bias from varying magnitudes and improving model convergence. Multiscale Entropic (MSE) analysis is employed for feature extraction, capturing both short- and long-term temporal patterns within transaction sequences, enhancing the representation of complex transactional behaviors. The extracted features are then processed using a Transformer-GAT classifier, which combines the attention mechanism of Transformers with Graph Attention Networks (GAT) to learn complex inter-transaction dependencies and graph-based relationships. This hybrid architecture enables the model to capture both local and global patterns, improving fraud detection performance. On the training dataset, the model achieved outstanding results with 98.65% accuracy, 98.70% precision, 98.50% recall, and an F1-score of 98.60 %, demonstrating a strong balance between correctly identifying fraudulent transactions and minimizing false alarms. The approach offers significant advantages for FinTech applications, including robust handling of imbalanced data, effective detection of subtle fraud patterns, and strong generalization to unseen transactions. 2025 IEEE. -
A Hybrid Framework for Detecting Hallucinations in Large Language Model Outputs
With Large Language Models (LLMs) continuously growing, they are on a path to replace the search engines soon. No matter how powerful they get, there is a certain level of uncertainty because they tend to hallucinate. Hallucination here refers to generate factually incorrect data, this can include making up names, generating false links and fabricating stories. This makes it extremely difficult to trust large language models. Existing papers provide solutions which are either not monetarily feasible or lack capabilities to build a robust hallucination detector. This paper aims to build a low resource hallucination detector which combines multiple heuristic signals like semantic similarity, self-consistency, external grounding via Wikipedia, NER overlap, flexible numerical check and a quantized LLM Falcon-7b. This eliminates the need to train the model from scratch. Upon evaluating with an input dataset of 50 questions the model was able to achieve an accuracy of 88%. 2025 IEEE. -
A Hybrid Genetic Algorithm and Large Language Model Approach for Agricultural Products Price Optimization
This paper introduces a hybrid approach, based on Genetic Algorithm (GA) and Large Language Models (LLMs), namely Mixtral 8x7B, to optimize pricing strategies for agricultural products. The method processes real-time market data, using Machine Learning (ML) techniques to generate competitive and profitable price recommendations. GAs allow for adaptive optimization, while LLMs capture complex trends in the market, making this approach more precise with respect to the pricing strategy. Case studies related to onions and tomatoes illustrate the efficiency of the optimization process. The outcome shows that the optimized prices achieve a fitness score of 0.915 (onions) and a competitive index of 0.89 (onions) compared to the market averages. Compared to traditional methods, the proposed hybrid model provides a better approach towards decision making through multi-objective optimization and real-time data analysis. This research contributes to improved profitability for farmers by adopting sustainable pricing strategies and agricultural market efficiency. 2025 IEEE. -
A hybrid GNNvanilla vision transformer model for IoT-based soil and crop forecasting
In this work, we propose a Graph?Neural Network (GNN) and Vanilla Transformer-based hybrid model for IoT driven soil and crop prediction. Conventional forecasting approaches are unable to model complicated spatial and temporal inter-dependencies and are not?very effective. The given paper solves this problem by using GNNs to learn the spatial relationships among the IoT sensor nodes and vanilla transformer model to?learn the temporal dependencies in crop and weather data. Vanilla vision transformer is able to recover missing contextual information during training. It is trained on data from IoT sensors that monitor soil moisture, temperature, humidity and a variety of other environmental factors as?well as historical crop yield and weather related information. The hybrid model can enable the real-time accurate prediction for crop?yield production and soil health status, which enables a smarter agriculture decision. The experimental results show that the proposed work achieves the lowest root mean square error (RMSE 2.1) and the highest crop accuracy (92%) for?short-term and long-term forecasts. Bharati Vidyapeeth's Institute of Computer Applications and Management 2025. -
A Hybrid Grayscale Image Scrambling Framework Using Block Minimization and Arnold Transform
Image disarranging is the process of randomly rearranging picture elements to make the visibility unreadable and break the link among neighboring elements. Pixel values often don't change while they are being scrambled. There has been a slew of proposed image encryption techniques recently. The two steps that most image encryption algorithms go through are confusion and diffusion. Using a scrambling technique, the pixel positions are permuted during the confusion phase, and an inverse-able function is used to modify the pixel values during the diffusion phase. A good scrambling method practically eliminates the high relationships between adjacent pixels in a picture. In the proposed scheme, XOR based minimization operator is applied on blocks of images followed by Arnold Transform. The suggested design is assessed using a matrix comprising the Structured Similarity Index and the Peak Signal to Noise Ratio. The computed PSNR value less than 10 indicates the input image and scrambled image has high variation. The SSIM value nearer to 0 indicates no similarity in the structure of the input image and scrambled image. 2024 IEEE. -
A Hybrid Intrusion Detection System for detecting Cross-layer DoS attacks in IoT
The Internet of Things (IoT) is critically prone to Denial of Service (DoS) attacks at multiple layers. If designed carefully, intrusion detection systems (IDS) can detect these attacks effectively. In the proposed study, we develop a Hybrid IDS to detect Cross-Layer DoS attacks in IoT. The proposed Cross-Layer system reduces the false positive rate considerably than a single IDS. The IDS is designed by ensembling multiple machine learning techniques to avoid overfitting or underfitting. The Hybrid IDS works in two stages, the first stage for detection of the attack occurrence (Anomaly detection) followed by a second stage to classify the attack types (Signature of the attacks). The output of the first stage is Correctly Detected Samples (CDS), which are again tested by the second stage to get Correctly Classified Samples (CCS). Another unique aspect of the proposed study is the dataset generation for different attacks considered. Rather than using the existing dataset, we have developed a trace file in NetSim Simulator by designing an attack environment. At the same time, during the feature selection process, a novel and efficient technique is applied to select the best feature set along with the critical component (CF). Simulation results accurately detect CDS of up to 95% and CCS of up to 96% with a weighted average F1 score. The testing time of the proposed model is also considerably lower than that of individual models, which makes the system efficient and lightweight. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025. -
A hybrid level set based approach for surface water delineation using landsat-8 multispectral images
The detection and delineation of surface water is a crucial step in change detection studies on water bodies using satellite images. Single band methods, spectral index methods, classification using machine learning and spectral un-mixing methods are the widely used strategies for surface water mapping from multi-spectral images. Level set theory based algorithms have been successfully employed in image segmentation problems and are proven to be effective. This study presents a hybrid level set theory based segmentation algorithm which is a combination of edge based and region based approaches to detect and delineate surface water bodies in Landsat 8 images. Level set algorithms were applied in combination with Modified Normalized Difference Water Index (MNDWI) to further improve the delineation accuracy. Robustness of the proposed approach was established by successfully applying the algorithm to delineate water bodies of different sizes, ranging from 0.5 km2 to 298 km2 in surface area. The proposed algorithm was also compared with established machine learning based delineation methods and found to be faster than the algorithms those produced comparable delineation outputs. As the ground truth was not available for accuracy measurement, the output image of the proposed method was compared with the outputs of the machine learning algorithms using Pearsons correlation co-efficient, Structural Similarity Index (SSIM) and Dice Similarity Index. The proposed algorithm was subsequently applied to multi-temporal Landsat data for water body change detection and analysis. 2021, International Association of Engineers. All rights reserved. -
A hybrid level shifted carrier-based PWM technique for modular multilevel converters
This paper presents a hybrid level shifted carrier-based pulse width modulation (HLSC-PWM) technique for modular multilevel converters (MMCs). The concept of the proposed HLSC-PWM method is developed by combining the principles of phase disposition PWM (PD-PWM), phase opposition disposition PWM (POD-PWM), and alternate phase opposition disposition PWM (APOD-PWM) methods. The main aim of the proposed HLSC-PWM method is to generate an output voltage with half-wave and quarter-wave symmetries. The generated symmetrical PWM output voltage based on the proposed HLSC-PWM method provides less total harmonic distortion (THD) and enhances the DC-Link voltage utilization (DCLVU). A generalized mathematical model is formulated to develop a single HLSC for MMC with an N number of submodules (SMs) per arm. Theoretical analysis of DCLVU for the proposed method is described. The functionality and performance of the HLSC-PWM method are carried out on a three-phase five-level MMC in MATLAB/Simulink. A hardware prototype of a single-phase five-level MMC is designed for experimental validation. The proposed HLSC-PWM method is implemented on an Altera/Cyclone I series (EP1C12Q240C8N) field-programmable gate array (FPGA), simulation and experimental results are presented. 2021 The Authors. IET Power Electronics published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology -
A Hybrid Machine Learning Model (NB-SVM) for Cardiovascular Disease Prediction
One of the leading causes of death is heart disease. The prediction of cardiovascular disease remains as a significant challenge in the clinical data analysis domain. Although predicting cardiac disease with a high degree of accuracy is highly challenging, it is possible with Machine Learning (ML) approaches. The implementation of an effective ML system can minimize the need for additional medical testing, minimize human intervention, and predict cardiovascular diseases with high accuracy. This type of assessment can reduce the disease's severity and mortality rate. Only a few studies show how machine learning techniques might forecast cardiac disease. This study presents a method for improving cardiovascular disease prediction accuracy using Machine Learning (ML) technologies. Various feature combinations and many known classification techniques are used to develop various cardio vascular disease prediction models. The proposed hybrid Machine Learning (ML) prediction model for heart disease leverages a higher degree of performance and accuracy. 2023 IEEE. -
A hybrid multi-optimizer approach using CNN and GB for accurate prediction of citrus fruit diseases
Efficient prediction of citrus fruit diseases is essential for maintaining orchard health and productivity. Traditional diagnostic methods, often relying on manual inspection, are labor-intensive and prone to inaccuracies. Deep learning techniques, especially Convolutional Neural Networks (CNNs), offer an automated and accurate alternative. This study introduces a novel model integrating CNN with Gradient Boosting (GB) and optimized using the Nesterov-Accelerated Adaptive Moment Estimation (Nadam) optimizer to enhance prediction accuracy. The model employs a custom CNN architecture combined with GB, leveraging Nadam for faster convergence and improved performance. Trained on a dataset of 3,000 citrus fruit images sourced from Kaggle, the model follows a structured process of preprocessing, feature extraction, integration of GB with CNN, and optimal prediction. Comparative analysis using metrics such as accuracy, precision, F1 score, and recall demonstrates the model's effectiveness, achieving an accuracy of 98.03% and precision of 98.04%. This robust approach addresses limitations of traditional methods by enabling automated feature extraction and reliable disease prediction. The proposed CNN-GB-Nadam model significantly enhances efficiency and reliability, providing a valuable tool for protecting citrus fruit health and improving orchard management practices. The Author(s) 2025.