Almaarefa University
The authoritative source for research information and outputs is Elsevier Scopus, with researcher profile verification from the Deanship of Scientific Research.

المخرجات

Research outputs

Articles, reviews, conference papers and book chapters, indexed by DOI and linked to University researchers.

14 output Results for «overfitting»

ReviewOpen access2026

AI-enabled eye-movement and emerging multimodal frameworks for precision dyslexia screening and reading pattern analysis

Frontiers in Medicine · vol. 13 · 1847464-1847464

Introduction: Developmental dyslexia refers to a common neurodevelopmental disorder, which impairs the accuracy and fluency of reading, and early identification is vital for initiating timely intervention. Nonetheless, the traditional methods of formal assessment are time- and resource-intensive, which limits their scalability. Machine-learning approaches and eye-tracking technologies provide objective, data-driven solutions for dyslexia screening. This research integrates current evidence on eye-movement-based and emerging multimodal computational methods for dyslexia screening, risk identification, and algorithmic classification during reading tasks. Methods: PubMed, Scopus, Web of Science, and CINAHL were searched systematically to identify studies published between January 2015 and March 2026. Eligible studies included analysis of eye-movement obtained via eye tracking or electrooculography (EOG), with or without predictive modeling. Methodological quality was assessed using JBI, PROBAST, ROBINS-I, and COSMIN tools. Results: = 2). The sample sizes ranged from small experimental cohorts (<20 participants) to larger datasets (>300 participants). In the literature, dyslexic readers were consistently found to exhibit longer fixation durations, increased regression behavior and reduced saccadic efficiency. Machine-learning algorithms using fixation, saccade, scan path, and signal-based features demonstrated classification accuracies ranging from approximately 80 to 95% with some studies reporting values approaching 99% under specific experimental conditions. Discussion: Nevertheless, there was a significant heterogeneity in datasets, feature extraction methods, outcome definitions and validation schemes. Notably, numerous studies used proxy diagnostic labels, small or internally derived datasets, and internal cross-validation, which introduces the risk of overfitting and performance inflation. Explicit multimodal or multi-source modeling was identified in three of 23 studies involving combinations of gaze data with demographic, cognitive, linguistic, VR-bed, text-derived, saliency-map, or CNN-based features. Two additional studies used EOG as an alternative eye-movement signal modality rather than true multi-source fusion. Therefore, the evidence base remains dominated by eye-movement and gaze-derived approaches, while multimodal evidence should be interpreted as emerging and exploratory. Altogether, eye-movement based computational systems are a promising, non-invasive method for scalable dyslexia screening. Systematic review registration: PROSPERO, identifier (RD42061332527).

developmental dyslexiadyslexia screeningeye movementseye tracking
ArticleOpen access202533citation· Semantic Scholar

Multi-Stage Neural Network-Based Ensemble Learning Approach for Wheat Leaf Disease Classification

IEEE Access · vol. 13 · 30101-30116

Prompt and accurate classification of wheat leaf diseases and their severity is essential for precise diagnosis, effective pesticide application, efficient disease control, and enhanced wheat production and quality. Nevertheless, the wide range of wheat diseases poses a significant challenge in terms of their identification, especially in intricate agricultural landscapes. The utilization of conventional models has major limitations in wheat disease detection, including dataset-specific performance, overfitting due to limited data, and high computational needs, making deployment in resource-constrained situations difficult. To address these challenges, this research proposes a multi-stage Convolutional Neural Network (CNN) based Ensemble Learning (EL) approach, which utilizes several concurrent CNN models with a bagged EL method to classify wheat leaf diseases. The various stages in the EL approach employ a multi-level framework to enhance feature extraction and capture complex data patterns. This improves the model’s emphasis on diseases while reducing the influence of complex backgrounds on disease identification. The proposed method, integrating pretrained CNNs and a bagging ensemble technique, achieved an accuracy of 86.78%, 98.28%, and 99.16% on the three publicly available datasets, outperforming the state-of-the-art models. These results demonstrate the potential of the proposed model for real-time disease diagnosis.

ensemble learningbaggingdisease severity assessmentplant disease diagnosis
ArticleOpen access202529citation· Semantic Scholar

Optimizing IoT intrusion detection with cosine similarity based dataset balancing and hybrid deep learning

Scientific Reports · vol. 15 (1) · 30939-30939

With IoT networks expected to exceed 29 billion connected devices by 2030, the risk of cyberattacks has never been higher. As more devices come online, the attack surface for hackers continues to expand, making cybersecurity a pressing concern. Intrusion Detection Systems (IDS) are essential for identifying and mitigating these threats in real-time. However, a significant challenge IDS faces is dealing with imbalanced datasets, where attack instances are significantly underrepresented compared to normal traffic. Training models on such skewed data leads to a bias toward majority-class patterns, reducing their ability to detect intrusions effectively. To address this issue, this work introduces CSMCR (Cosine Similarity-based Majority Class Reduction), a novel technique that selectively removes redundant majority-class samples while preserving dataset integrity. Unlike traditional approaches like SMOTE (oversampling) or random undersampling, CSMCR ensures that the retained majority instances remain diverse by analyzing feature-wise similarity. This prevents unnecessary data duplication and minimizes information loss. Additionally, we developed a hybrid deep learning model integrating RegNet and FBNet architectures to enhance feature extraction and classification performance. Experimental results on multiple IDS datasets confirm that balancing the dataset to a 1:1 ratio optimally prevents overfitting and improves model interpretability. The proposed model achieved an F1-score of 0.9758 on RT-IoT2022 and 0.9275 on UNSW Bot-IoT, outperforming SMOTE-based methods in accuracy and computational efficiency. Notably, CSMCR reduced training time by 53% compared to conventional oversampling techniques. Incremental training evaluations reveal that bias formation reduces performance beyond a 1:2 majority-to-minority ratio. These findings establish CSMCR as a robust, scalable, and computationally efficient IDS balancing strategy tailored for IoT network security.

IoT intrusion detectionclass imbalancecosine similaritydata balancing
ArticleOpen access20258citation· Semantic Scholar

An ideally designed deep trust network model for heart disease prediction based on seagull optimization and Ruzzo Tompa algorithm

Scientific Reports · vol. 15 (1) · 6035-6035

Diet, stress, genetics, and a sedentary lifestyle may all contribute to heart disease rates. Although recent studies propose comprehensive automated diagnostic systems, these systems tend to focus on one aspect, such as feature selection, prioritization, or predictive accuracy. A more complete approach that considers all of these factors can improve the efficiency of a cardiac prediction system. This study uses an appropriate strategy to overcome potential network design problems, design challenges, overfitting, and lack of robustness that can interfere with system performance. The research introduces an ideally designed deep trust network called ID-DTN to improve system performance. The Ruzzo-Tompa method is used to eliminate noncontributory features. The Seagull Optimization Algorithm (SOA) is introduced to optimize the trust depth network to achieve optimal network design. The study scrutinizes the deep trust network (ID-DTN) and the restricted Boltzmann machine (RBM) and sheds light on the system's operation. This proposal can optimize both network architecture and feature selection, which is the main novelty. The proposed method is analyzed using the below-mentioned metrics: Matthew's correlation coefficient, F1 score, accuracy, sensitivity, specificity, and accuracy. ID-DTN performs well compared to other state-of-the-art methods. The validation results confirm that the proposed method improves the prediction accuracy to 97.11% and provides reliable recommendations for patients with cardiovascular disease.

heart disease predictionseagull optimization algorithmfeature selectionnetwork structure optimization
Conference paperClosed20252citation· Semantic Scholar

Advanced Techniques for Self Supervised Learning in Medical Imaging Using Multi Instance Contrastive Learning

OPJU International Technology Conference on Smart Computing for Innovation and Advancement in Industry (OTCON) · 1-6

Advanced self-supervised learning approaches in medical imaging improve diagnosis accuracy and patient health. This research focuses on multi-instance contrastive learning approaches that employ unstructured data to produce usable representations. We examine DINO, DeepCluster, CMC, InfoNCE, and others. All of them employ contrastive learning to discover patterns in complex medical datasets. The research emphasizes the need of making positive samples more similar and negative samples less similar to address medical imaging data's various difficulties. Dynamic thresholding, rigorous negative mining, and regularization reduce overfitting and improve model usefulness. Evaluate-based evolution ensures model performance improves over time. The recommended strategy outperforms existing methods on certain performance assessment parameters. It demonstrates how these strategies could potentially change professional procedures. As medical imaging improves, more individuals realize self-supervised learning may transform diagnosis and treatment. This research shows how innovative data analysis may improve medical imaging, benefiting physicians and patients.

self-supervised learningdiagnostic imagingDINOInfoNCE
ArticleOpen access20252citation· Semantic Scholar

Revolutionizing ALS Assessment: XGBoost Classification with Progressive Entropy Weighted-based Focal Loss on Gene Sequences

Journal of Disability Research · vol. 4 (1)

In recent times, the global rise in the prevalence rate of amyotrophic lateral sclerosis (ALS) has profoundly affected the welfare of several people in the world. ALS is a lethal neurodegenerative disease (NDD) that damages the nerve cells in the brain and spinal cord. Moreover, it removes the person’s capability of controlling muscle movements in the body. It is necessary to detect the disease earlier, to reduce the disease severity, and to enhance the life expectancy of the patients. Traditionally, ALS screening is handled by qualified physicians through blood tests, which is an expensive, painful, and time-consuming process. To resolve this limitation, several researchers focused on the ALS classification. Conversely, it have a few drawbacks, such as lack of accuracy and speed, overfitting of data, and noise handling tasks. For enhancing the classification of ALS, the proposed approach employs progressive entropy weighted-based focal loss (PEWFL)-XGBoost through the Kaggle ALS dataset. The XGBoost is used for the ability to manage missing data and speed. Nevertheless, it has certain limitations such as overfitting of data, hyperparameter tuning, and handling of smaller datasets. To resolve this, PEWFL is added to the XGBoost system to improve the classification performance. Correspondingly, the efficiency of the respective system is calculated using performance metrics to evaluate the performance of the research. Moreover, internal comparison with classical algorithms such as XGBoost, K-nearest neighbor, and random forest discloses the efficacy of the respective model. The proposed system is envisioned to contribute to molecular genetics and neuroscience research and assist neurologists in enhancing the diagnosis of ALS.

amyotrophic lateral sclerosisgene sequenceXGBoostneurodegenerative disease diagnosis
ArticleOpen access202416citation· Semantic Scholar

Group Teaching Optimization With Deep Learning-Driven Osteosarcoma Detection Using Histopathological Images

IEEE Access · vol. 12 · 34089-34098

Osteosarcoma is the most normal kind of cancer that arises in bones, which appears on the surface to resemble earlier types of bone cells that assist in forging new bone tissues, but the tissue in osteosarcoma is weaker and softer than normal bone tissue. The usage of automated techniques for the detection of osteosarcoma has the potential to mitigate the obligations and burdens confronted by pathologists owing to its abundant quantity of cases. Artificial intelligence (AI) has an emerging progress in diagnostic pathology. In recent years, numerous studies using deep learning (DL) techniques to histopathological images (HI) have been published. While several studies claim higher accuracy, they might lack generalization and fall into the pitfall of overfitting owing to the wide range of HI. The study objective is to enhance the diagnosis and detection of osteosarcoma by employing computer-assisted detection (CAD) and diagnoses (CADx). Technique like convolutional neural networks (CNN) make better prognoses for patient conditions and considerably reduce the surgeon’s workload. CNN needs to be trained on the massive quantity of data to accomplish a remarkable performance. Therefore, the study presents a novel Group Teaching Optimization Algorithm with Deep Learning-Driven Osteosarcoma Detection on Histopathological Images (GTOADL-ODHI) technique. The purpose of the GTOADL-ODHI technique is to examine the HIs for the detection and classification of osteosarcoma. To accomplish this, the GTOADL-ODHI algorithm applies the Gaussian filtering (GF) method for image pre-processed to become rid of the noise. Besides, the capsule network (CapsNet) model is utilized for the extractor of the feature vector. Furthermore, the hyperparameter selection of the CapsNet approach takes place using the GTOA. Finally, the self-attention bidirectional long short-term memory (SA-BiLSTM) model can be employed for osteosarcoma recognition and classification. The widespread experimental analysis of the GTOADL-ODHI method is tested on the benchmark datasets. The simulation validation reported the optimum solution of the GTOADL-ODHI algorithm related to existing systems concerning distinct aspects.

histopathological imagescapsule networkshyperparameter optimizationcomputer-aided diagnosis
ArticleOpen access20248citation· Semantic Scholar

A Fine-Tuned CatBoost-Based Speech Disorder Detection Model

Journal of Disability Research · vol. 3 (3)

The classification of speech disorders (SDs) is crucial for treating children with speech impairment (SI). An automated SD classification can assist speech therapists in rendering services to children with SI in rural areas. Automated techniques for detecting SDs provide objective assessments of speech attributes, including articulation, fluency, and prosody. Clinical examinations and quantitative assessments provide an in-depth understanding of the patient’s speaking abilities and limitations. Existing deep learning (DL) models for SD detection often lack generalization across diverse populations and speech variations, leading to suboptimal performance when applied to individuals with different linguistic backgrounds or dialects. This study introduces a DL-based model for classifying normal and abnormal speeches using voice samples. To overcome the overfitting and bias, the authors construct convolutional neural network models with the weights of MobileNet V3 and EfficientNet B7 models for feature extraction (FE). To improve performance, they integrate the squeeze and excitation block with the MobileNet V3-based FE model. Similarly, the EfficientNet B7-model-based FE is improved using the structure pruning technique. The enhanced CatBoost model differentiates the normal and abnormal speeches using the extracted features. The experimental analysis is performed using the public dataset that contains 4620 utterances of healthy children and 2178 utterances of children with SI. The comparative study reveals the exceptional performance of the proposed SD classification model. The model outperforms the current SD classification models. It can be employed in clinical settings to support speech therapists. Substantial training with diverse voice samples can improve the generalizability of the proposed model.

CatBoostconvolutional neural network feature extractionMobileNetV3EfficientNetB7
ArticleOpen access20241citation· Semantic Scholar

Trailblazing ASD Behavioral Research: Leveraging Modified MLP with Cross-Weighted Attention Mechanism at the Forefront

Journal of Disability Research · vol. 3 (4)

Autism spectrum disorder (ASD) is an enduring and intricate condition that encompasses issues with behavior and communication. This disability has become common among several individuals globally. To ensure that all the autistic individuals get the correct support and care throughout their lifetime, it is significant that the autism signs are realized and the condition is spotted in the initial phase itself to avoid further complications. However, the process of ASD diagnosis is time-consuming, with expensive testing procedures. An effective screening method is needed for early treatment to improve the quality of the ASD patient’s life. Several conventional studies thrived on attaining ASD identification with machine learning and deep learning) but lacked in accuracy and computation. Therefore, the propounded system employs modified multi-layer perceptron (MLP) with cross-weighted attention mechanism. MLP is utilized for its capability in resolving intricate nonlinear complications, handling huge datasets, and enhancing the accuracy of the model. Though existing studies have utilized MLP for various identification processes, it still lacks identification speed, has overfitting issues, and requires more parameters. To overcome these issues, the proposed system employs cross-weighted attention mechanism, to enhance the identification process. As few researchers focused only on identifying and diagnosing the ASD adult, the proposed system uses autism screening in adult datasets and intends to improve the diagnosis of ASD adult patients. Attention ASD-modified MLP with cross-weighted algorithm is applied to classify and perform with various algorithms such as random forest, MLP, and Naïve Bayes. Furthermore, the performance is examined with certain metrics to calculate the efficacy of the proposed system.

autism spectrum disorderautism screeningadult autism diagnosisearly diagnosis
ArticleOpen access20241citation· Semantic Scholar

Hybrid Feature Extraction Technique-based Alzheimer’s Disease Detection Model Using MRI Images

Journal of Disability Research · vol. 3 (6)

Detecting Alzheimer’s disease (AD) using magnetic resonance imaging (MRI) is essential for early diagnosis and management. This study introduces a new method for detecting AD by combining three robust models: DenseNet201, EfficientNet B7, and extremely randomized trees (ERT). We improve the ability to extract features in DenseNet201 by including a self-attention mechanism. Additionally, we use early stopping techniques on EfficientNet B7 to address the issue of overfitting. In addition, Bayesian Optimization and Hyperband optimization techniques are used to adjust the hyperparameters of extra-trees to differentiate normal and abnormal MRI images. In addition, the authors used SHapley Additive exPlanations to understand the model’s decision. With minimal computer resources, the proposed model achieved a remarkable accuracy of 98.9% in detecting AD. The findings highlight the effectiveness of recommended feature extraction and ERT models and optimization methods to accurately identify AD using MRI images.

Alzheimer's disease detectionMRI imageshybrid feature extractionDenseNet201
ArticleOpen access202354citation· Semantic Scholar

Hybrid model for precise hepatitis-C classification using improved random forest and SVM method

Scientific Reports · vol. 13 (1) · 12473-12473

Hepatitis C Virus (HCV) is a viral infection that causes liver inflammation. Annually, approximately 3.4 million cases of HCV are reported worldwide. A diagnosis of HCV in earlier stages helps to save lives. In the HCV review, the authors used a single ML-based prediction model in the current research, which encounters several issues, i.e., poor accuracy, data imbalance, and overfitting. This research proposed a Hybrid Predictive Model (HPM) based on an improved random forest and support vector machine to overcome existing research limitations. The proposed model improves a random forest method by adding a bootstrapping approach. The existing RF method is enhanced by adding a bootstrapping process, which helps eliminate the tree's minor features iteratively to build a strong forest. It improves the performance of the HPM model. The proposed HPM model utilizes a 'Ranker method' to rank the dataset features and applies an IRF with SVM, selecting higher-ranked feature elements to build the prediction model. This research uses the online HCV dataset from UCI to measure the proposed model's performance. The dataset is highly imbalanced; to deal with this issue, we utilized the synthetic minority over-sampling technique (SMOTE). This research performs two experiments. The first experiment is based on data splitting methods, K-fold cross-validation, and training: testing-based splitting. The proposed method achieved an accuracy of 95.89% for k = 5 and 96.29% for k = 10; for the training and testing-based split, the proposed method achieved 91.24% for 80:20 and 92.39% for 70:30, which is the best compared to the existing SVM, MARS, RF, DT, and BGLM methods. In experiment 2, the analysis is performed using feature selection (with SMOTE and without SMOTE). The proposed method achieves an accuracy of 41.541% without SMOTE and 96.82% with SMOTE-based feature selection, which is better than existing ML methods. The experimental results prove the importance of feature selection to achieve higher accuracy in HCV research.

hepatitis C virushybrid predictive modelimproved random forestbootstrapping
ArticleOpen access2022364citation· Semantic Scholar

A Comparison of Pooling Methods for Convolutional Neural Networks

Applied Sciences · vol. 12 (17) · 8643-8643

One of the most promising techniques used in various sciences is deep neural networks (DNNs). A special type of DNN called a convolutional neural network (CNN) consists of several convolutional layers, each preceded by an activation function and a pooling layer. The feature map of the previous layer is sampled by the pooling layer (that seems to be an important layer) to create a new feature map with condensed resolution. This layer significantly reduces the spatial dimension of the input. It always accomplished two main goals. As a first step, it reduces the number of parameters or weights to minimize computational costs. The second step is to prevent the overfitting of the network. In addition, pooling techniques can significantly reduce model training time and computational costs. This paper provides a critical understanding of traditional and modern pooling techniques and highlights the strengths and weaknesses for readers. Moreover, the performance of pooling techniques on different datasets is qualitatively evaluated and reviewed. This study is expected to contribute to a comprehensive understanding of the importance of CNNs and pooling techniques in computer vision challenges.

pooling layersconvolutional neural networksoverfitting reductioncomputational complexity reduction
ArticleOpen access202220citation· Semantic Scholar

Multi-View Clustering Based on Multiple Manifold Regularized Non-Negative Sparse Matrix Factorization

IEEE Access · vol. 10 · 113249-113259

Clustering of multi-view data has got broad consideration of the researchers. Multi-view data is composed through different domain which shows the consistent and complementary behavior. The existing studies did not draw attention of over-fitting and sparsity among the diverse view, which is the considerable issue for getting the unique consensus knowledge from these complementary data. Herein article, a multi-view clustering approach is recommended to provide the consensus solution from the multi-view data. To accomplish this task, we exploit non-negative matrix factorized method to generate a cost function. Further, manifold learning model is used to build the graph through the nearest neighbor strategy, which is effective to save the geometrical design for data and feature matrix. Furthermore, the over-fitting problem, sparsity is handled through adaption of frobenious norm, and$L_{1}$-norm on basis and coefficient matrices. The whole formulation is done through the mathematical function, which is optimized through the iterative updating strategy to get the optimal solution. The computational experiment is carried on the available datasets to exhibits that the proposed strategy beats the current methodologies in terms of clustering execution.

multi-view clusteringmanifold regularizationconsensus clusteringoverfitting
Conference paperClosed202135citation· Semantic Scholar

Machine LearningTechniquesfor Detection of Website Phishing: A Review for Promises and Challenges

IEEE Annual Computing and Communication Workshop and Conference (CCWC) · 0813-0818

Websites phishing is a cyber-attack that targets online users to steal their sensitive informationincluding login credentials and banking details. Attackers fool the users by presenting the masked webpage as legitimate or trustworthy to retrieve their essential data. Several solutions to phishing websites attacks have been proposed such as heuristics, blacklist or whitelist, and Machine Learning (ML) based techniques. This paper presents the state of art techniques for phishing website detection using the ML techniques. This research identifies solutions to the website's phishing problem based on the ML techniques. The majority of the examined approaches are focused on traditional ML techniques. Random Forest (RF), Support Vector Machine (SVM), Naïve Bayes (NB), and Ada Boosting are the powerful ML techniques examined in the literature. This survey paper also identifies deep learning-based techniques with better performance for detecting phishing websites compared to the conventional ML techniques. Challenges to ML techniques identified in this work include overfitting, low accuracy, and ML techniques' ineffectiveness in case of unavailability of enough training data. This research suggests that Internet users should know about phishing to avoid cyber-attacks. This paper also points out the proposal for an automated solution to phishing websites.

phishing detectionphishing websitesmachine learningdeep learning