Ghaith M. Jaradat

This review examines recent advances and persistent challenges in multi-class classification within machine learning (ML) and deep learning (DL), a core task underpinning many real-world applications in healthcare, finance, social media, and other high-impact domains. The review provides a structured analytical synthesis of major methodological directions, including problem transformation methods, algorithm-level approaches, ensemble and hybrid strategies, class-imbalance handling, evaluation metrics, and deployment-related considerations such as interpretability, uncertainty, and ethics. Recent progress in deep learning architectures, ensemble learning, and active learning has substantially improved predictive accuracy, robustness, and data efficiency across diverse application settings. At the same time, important challenges remain, including imbalanced datasets, noisy labels, scalability constraints, computational cost, and the need for trustworthy and transparent decision-making. The review also examines the trade-offs among predictive performance, interpretability, fairness, and deployment feasibility, highlighting the importance of selecting methods and evaluation metrics that are appropriate to application context. Rather than proposing a new algorithm, this work offers an integrative framework that connects technical advances with practical and ethical considerations in multi-class classification. Finally, it identifies key future directions, including semi-supervised and transfer learning, few-shot and federated multi-class systems, robustness under label noise, energy-efficient model design, and scalable interpretability frameworks for high-stakes deployment.

The early and accurate prediction of diabetes mellitus remains a significant challenge in clinical decision-making due to the high dimensionality, noise, and heterogeneity of medical data. This study proposes a novel hybrid classification framework that integrates the dwarf mongoose optimization (DMO) algorithm for feature selection with a convolutional neural network–long short-term memory (CNN-LSTM) deep learning architecture for predictive modeling. The DMO algorithm is employed to intelligently select the most informative subset of features from a large-scale diabetes dataset collected from 130 U.S. hospitals over a 10-year period. These optimized features are then processed by the CNN-LSTM model, which combines spatial pattern recognition and temporal sequence learning to enhance predictive accuracy. Extensive experiments were conducted and compared against traditional machine learning models (logistic regression, random forest, XGBoost), baseline deep learning models (MLP, standalone CNN, standalone LSTM), and state-of-the-art hybrid classifiers. The proposed DMO-CNN-LSTM model achieved the highest classification performance with an accuracy of 96.1%, F1-score of 94.6%, and ROC-AUC of 0.96, significantly outperforming other models. Additional analyses, including confusion matrix, ROC curves, training convergence plots, and statistical evaluations confirm the robustness and generalizability of the approach. These findings suggest that the DMO-CNN-LSTM framework offers a powerful and interpretable tool for intelligent diabetes prediction, with strong potential for integration into real-world clinical decision-support systems.

Sentiment analysis (SA) is a vital text-mining task for extracting and classifying opinions from textual data as positive, negative, or neutral. This research investigates sentiment analysis for the Jordanian Arabic dialect by integrating a Memetic Algorithm (MA) for feature selection with three machine learning classifiers: Support Vector Machine (SVM), Multinomial Naïve Bayes (MultinomialNB), and Logistic Regression (LR). The core strategy of the MA is to combine global and local search to filter out irrelevant features and select the most relevant ones, thereby optimising the feature space for classification. The model's accuracy served as the fitness function for the global search (genetic algorithm), while Pearson's correlation coefficient was used for the local search. Experimental results on the Arabic Jordanian General Tweets (AJGT) dataset demonstrated that the MA consistently improved the performance of all classifiers. The baseline accuracies were SVM (79.11%), MultinomialNB (83.89%), and LR (82.67%). After applying the MA, the accuracies improved to SVM (82.22%), MultinomialNB (85.61%), and LR (84.56%). Similar enhancements were observed in precision, recall, and F1-score across all models. The study concludes that MultinomialNB, when coupled with the MA, is the most effective classifier for this task, and the hybrid MA approach significantly boosts sentiment classification performance for the Jordanian dialect.

The acceleration of digital transformation in Saudi Arabia, driven by the Vision 2030 initiative, has significantly expanded the use of Mobile Tourism Services Applications (MTSAs). Ensuring their technical quality is essential to enhancing both user satisfaction and service delivery. This study introduces a tailored Technical Quality Model for MTSAs that incorporates usability, performance, security, and satisfaction, while also accounting for the cultural, regulatory, and infrastructural context of Saudi Arabia. Adopting a mixed-methods design, the research integrates quantitative responses from 500 users and 50 developers with qualitative evidence from interviews and document analysis. Statistical findings highlight security as the strongest determinant of user satisfaction (mean = 4.5, SD = 0.7; Pearson r = 0.75). Regression analysis further indicates that security (β = 0.52, p < 0.001) explains 65% of the variance in satisfaction (R² = 0.65). Gender and experience were found as significant moderators, with males and more experienced users reporting higher satisfaction (t = 2.34, p = 0.02; t = 3.45, p = 0.001). It is concluded by recommending enhancements in usability, robust security mechanisms, and gender-responsive design to optimize user experiences. Future research should explore integrating artificial intelligence and blockchain, and validate the proposed model in comparable regional contexts.

Estimating the number of wind ranches generated in the upcoming minutes, hours, or days is the focus of wind power forecasting. Deep learning has garnered a lot of interest in wind control estimation because of how well they perform classification, grouping, and recurrence. The adaptive neuro-fuzzy inference system was successfully applied in wind power forecasting. However, its performance relies on optimal selection of hyperparameters. This study introduces a novel predictive model by incorporating the reptile search algorithm with adaptive neuro-fuzzy inference system (ANFIS) for short-term wind power forecasting. It employs reptile search algorithm (RSA), known for adjustable parameters, disentangled search, and consistent outcomes, to optimize ANFIS’s hyperparameters. Additionally, via exploitation during training, RSA performs a selection of best features in the dataset that contributes to the classification accuracy of ANFIS. This aims to enhance precision of the anticipated yield. Employing authentic wind power data from Jordan is undertaken to evaluate efficiency. The performance is compared with alternative techniques, including artificial neural networks, random forests, and support vector machines. Findings showed that ANFIS-RSA performs competitively for the well-known Chinese benchmark dataset (99.9% accuracy; 0.99 R2; 10.54 MAE; 11.62 RMSE) and is more robustly accurate than others over the Jordanian dataset (0.84.6% accuracy; 0.96 R2; 0.098 MAE; 0.203 RMSE).

The widespread use of social media has brought many challenges, mainly due to a misconstrued interpretation of the right to freedom of expression. Cyberbullying is a particularly noteworthy issue with far-reaching global implications for both its victims and the wider community. It takes the form of bullying that happens on several social media websites. This paper's goal is to develop a deep learning model capable of recognizing cases of cyberbullying on social media. Four models, such as bidirectional long short-term memory (BiLSTM), convolutional neural network with bidirectional long short-term memory (CNN-BiLSTM), bidirectional long short-term memory with gated recurrent unit (BiLSTM-GRU), and artificial neural network (ANN), will be evaluated in a multiclass classification difficulty context. The results showed that the BiLSTM model outperformed the other models by achieving the highest accuracy in 91% of cases, while the CNN-BiLSTM and ANN models demonstrated relatively lower performance. In addition to determining the efficacy of the deep learning techniques, the work highlights the urgent requirement for strong systems to resist cyberbullying. By enhancing detection accuracy, the proposed model can contribute significantly to providing a safer digital environment for further studies in this field. Received: 24 August 2024 | Revised: 16 December 2024 | Accepted: 25 February 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available on request from the corresponding author upon reasonable request. Author Contribution Statement Ghaith Jaradat: Conceptualization, Methodology, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Mohammad Shehab: Conceptualization, Methodology, Software, Validation, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Supervision, Project administration. Dyala Ibrahim: Software, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Project administration. Saif Najdawi: Methodology, Validation, Writing – original draft, Writing – review & editing, Visualization. Rami Sihwail: Software, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization.

In recent years, the extensive growth of the Internet, coupled with the integration of sensors and wireless sensor networks into critical areas like healthcare and military defense, has led to a significant expansion in the use of artificial intelligence and the Internet of Things applications. Due to the importance and sensitivity of the data in these fields, it is crucial to use an intrusion detection system (IDS), which applies specific algorithms to analyze and process the data from the network in order to find any suspicious activities or behaviors on the system and improve system security. A network frequently carries enormous amounts of data, particularly in military applications where data must flow continuously. In this research, an enhanced IDS using the scatter search algorithm is proposed. A population of randomly generated initial solutions is used to produce a diverse group of selected and high-performing solutions, which serve as a reference set to steer the search process. This reference set is then adopted as a feature selection method based on a support vector machine. The efficiency of the suggested method was examined utilizing the NSL-KDD dataset, and the results were compared with the Gazelle Optimization Algorithm, Algorithmic Optimization Algorithm, Gray Wolf Optimizer, Adjusted Gray Wolf Optimizer, and Particle Swarm Optimization. The main performance metrics utilized to test the efficiency of the suggested method include accuracy, detection efficiency, false-positive rate, and feature count. The results illustrated that the suggested method has obtained a high intrusion detection accuracy in the IDS system of 99% and decreased false alarm rates (0.02) and selected only 17 features from the initial dataset, which contained 41 features, demonstrating the effectiveness of the suggested method. Received: 6 October 2024 | Revised: 14 January 2025 | Accepted: 13 February 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data are available from the corresponding author upon reasonable request. Author Contribution Statement Feras E. AbuAladas: Conceptualization, Software, Validation, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Mohammad Shehab: Conceptualization, Validation, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Rasha Israwah: Methodology, Validation, Investigation, Data curation, Writing – original draft, Writing – review & editing. Ghaith Jaradat: Methodology, Formal analysis, Writing – original draft, Writing – review & editing, Visualization. Yousef Qawqzeh: Software, Investigation, Writing – original draft, Writing – review & editing.

The predictive directional greedy routing (PDGR) protocol is widely utilized in highway settings within vehicular ad hoc networks (VANETs). However, PDGR encounters a notable challenge when packets lack a suitable vehicle directionally, leading to network disconnections. This triggers a shift to carry and forward recovery mode due to outdated neighbor information in the vehicle's neighbor table (VNT). To address this, our study proposes an improved fuzzy logic-based improved PDGR (IPDGR). This novel algorithm dynamically adjusts beaconing intervals based on real-time network dynamics. Through comprehensive evaluation using VANET simulators, IPDGR demonstrates superior performance compared to PDGR and directional greedy routing (DGR) protocols across various metrics including Inconsistency of vehicle's neighbor's table (IVNT), packet delivery ratio (PDR), routing path length (RPL), and number of hole problem occurrence (NHPO).

Heart disease remains a leading cause of mortality worldwide, prompting healthcare researchers to leverage analytical tools for comprehensive data analysis. This study focuses on exploring crucial parameters and employing deep learning (DL) techniques to enhance understanding and prediction of cardiovascular disease (CVD) risk factors. Utilizing SPSS and Weka tools, a cross-sectional and correlational design was employed to analyze extensive medical datasets. Binomial regression analysis revealed significant associations between age (𝑝 = 0.004) and body mass index (𝑝 = 0.002) with CVD development, highlighting their importance as risk factors. Leveraging Weka's DL algorithms, a predictive model was constructed to classify CVD causes. Particularly, convolutional neural networks (CNN) showcased remarkable accuracy, reaching 98.64%. The findings underscore the elevated risk of CVD among university students and employees in Saudi Arabia, emphasizing the need for heightened awareness and preventive measures, including dietary improvements and increased physical activity. This study underscores the importance of further research to enhance CVD risk perception among students and individuals in similar settings.

In the healthcare field, diagnosing disease is the most concerning issue. Various diseases including cardiovascular diseases (CVDs) significantly influence illness or death. On the other hand, early and precise diagnosis of CVDs can decrease chances of death, resulting in a better and healthier life for patients. Researchers have used traditional machine learning (ML) techniques for CVD prediction and classification. However, many of them are inaccurate and time-consuming due to the unavailability of quality data including imbalanced samples, inefficient data preprocessing, and the existing selection criteria. These factors lead to an overfitting or bias issue towards a certain class label in the prediction model. Therefore, an intelligent system is needed which can accurately diagnose CVDs. We proposed an automated ML model for various kinds of CVD prediction and classification. Our prediction model consists of multiple steps. Firstly, a benchmark dataset is preprocessed using filter techniques. Secondly, a novel arithmetic optimization algorithm is implemented as a feature selection technique to select the best subset of features that influence the accuracy of the prediction model. Thirdly, a classification task is implemented using a multilayer perceptron neural network to classify the instances of the dataset into two class labels, determining whether they have a CVD or not. The proposed ML model is trained on the preprocessed data and then tested and validated. Furthermore, for the comparative analysis of the model, various performance evaluation metrics are calculated including overall accuracy, precision, recall, and F1-score. As a result, it has been observed that the proposed prediction model can achieve 88.89% accuracy, which is the highest in a comparison with the traditional ML techniques.