The rise of telemedicine platforms like Altibbi has transformed healthcare access, enabling patients to consult specialists remotely. One vital component of these platforms is medical question classification, which ensures that patient inquiries are efficiently directed to the appropriate specialists. However, these datasets often suffer from class imbalance, leading to biased model performance. This study explores the use of AraBERTv0.2-Twitter embeddings combined with ensemble learning techniques, Bagging and Boosting applied to various classifiers, including Logistic Regression (LR), Random Forests (RF), Decision Trees (DT), and K-Nearest Neighbors (KNN). We evaluate their performance on a dataset of Arabic medical questions from Altibbi under four experimental conditions: (1) Bagging without Synthetic Minority Over-sampling Technique (SMOTE), (2) Bagging with SMOTE, (3) Adaptive Boosting (AdaBoost) without SMOTE, and (4) AdaBoost with SMOTE. This allows us to analyze the impact of synthetic oversampling on classification performance and identify the most effective approach for addressing data imbalance. Experimental results show that Bagging with LR, without SMOTE, achieves the highest Macro F1-score (0.8468) and G-Mean (0.9077). AdaBoost with LR benefits significantly from oversampling, reaching a G-Mean of (0.8921) with SMOTE. Additionally, RF improves G-Mean from (0.8831) to (0.8893) when combined with Bagging and SMOTE. These findings highlight the vital role of ensemble methods in enhancing classification fairness and accuracy. By integrating such strategies, telemedicine platforms can significantly improve the automated classification of medical questions, ensuring patients receive timely and accurate responses from the right specialists.

The growing demand for telemedicine has highlighted the need for automated healthcare services, particularly in medical question classification. This study presents a deep learning model designed to address key challenges in telemedicine, including class imbalance and accurate routing of Arabic medical questions to the correct specialties. The model combines AraBERTv0.2-Twitter, fine-tuned for informal Arabic, with Bidirectional Long Short-Term Memory (BiLSTM) networks to capture deep semantic relationships in medical text. We used a labeled dataset of 5000 Arabic consultation records from Altibbi, covering five key medical specialties selected for their clinical relevance and frequency. The data underwent preprocessing to remove noise and normalize text. We employed stratified sampling to ensure representative distribution across the selected medical specialties. We evaluate multiple models using macro precision, macro recall, macro F1-score, weighted F1-score, and G-Mean. Our results demonstrate that DeepSMOTE combined with cross-entropy loss achieves the best performance. The findings offer statistically significant improvements and have practical implications for improving screening and patient routing in telemedicine platforms.

The detection of spam reviews in multilingual environments remains a challenging task due to linguistic diversity, data imbalance, and semantic complexity. This paper proposes a novel hybrid model that integrates Twin Support Vector Machine (TwinSVM) with Harris Hawks Optimization (HHO) for simultaneous parameter optimization and feature selection. To enhance semantic understanding, sentiment-based features are incorporated alongside pre-trained word embedding models-BERT, FastText, and MUSE-across English, Arabic, and Spanish datasets. Our approach generates 24 high-quality datasets using embeddings with 100 and 400 dimensions, including a combined multilingual set. Experimental results demonstrate that our proposed HHO-TwinSVM model consistently outperforms conventional classifiers and metaheuristic-enhanced SVMs, achieving accuracy improvements of up to 9.44% and enhanced robustness in low-resource languages. This integrated framework represents a scalable and adaptable solution for multilingual spam detection. Four detailed experiments were conducted in this study, each designed to address and demonstrate a specific aspect of the proposed approach. Across all experiments, the method outperformed existing algorithms, achieving impressive accuracy rates of 92.9741 %, 89.0314%, 80.3580%, and 85.0859% on Arabic, English, Spanish, and multilingual datasets, respectively. Subsequently, sentiment analysis features were incorporated to further enhance detection performance, resulting in improvements of 1.0994%, 2.6674%, 9.4430%, and 8.7448%, respectively. A comprehensive analysis of the experimental results, including the influence of reviews and sentiment features, is also presented.

As the business world shifts to the web and tremendous amounts of data become available on multilingual mobile applications, new business and research challenges and opportunities have been explored. This research aims to intensify the usage of data analytics, machine learning, and sentiment analysis of textual data to classify customers’ reviews, feedback, and ratings of businesses in Jordan’s food and restaurant industry. The main methods used in this research were sentiment polarity (to address the challenges posed by businesses to automatically apply text analysis) and bio-metric techniques (to systematically identify users’ emotional states, so reviews can be thoroughly understood). The research was extended to deal with reviews in Arabic, dialectic Arabic, and English, with the main focus on the Arabic language, as the application examined (Talabat) is based in Jordan. Arabic and English reviews were collected from the application, and a new model was proposed to sentimentally analyze reviews. The proposed model has four main stages: data collection, data preparation, model building, and model evaluation. The main purpose of this research is to study the problem expressed above using a model of ordinal regression to overcome issues related to misclassification. Additionally, an automatic multi-language prediction approach for online restaurant reviews was proposed by combining the eXtreme gradient boosting (XGBoost) and particle swarm optimization (PSO) techniques for the ordinal regression of these reviews. The proposed PSO-XGB algorithm showed superior results when compared to support vector machine (SVM) and other optimization methods in terms of root mean square error (RMSE) for the English and Arabic datasets. Specifically, for the Arabic dataset, PSO-XGB achieved an RMSE value of 0.7722, whereas PSO-SVM achieved an RSME value of 0.9988.

Random Vector Functional Link (RVFL) networks have garnered attention as a rapid and efficient neural network model due to their simplified architecture and reduced training complexity. Nevertheless, the hyperparameter tuning of this network remains a substantial obstacle in the pursuit of enhanced performance across many applications. In this study, we present GrafoRVFL, an open-source framework that employs gradient-free algorithms to optimize RVFL networks' hyperparameters. GrafoRVFL is a system that is adaptable and helps to enhance the performance of RVFL models. It is constructed on top of Numpy, Mealpy, and Scikit-Learn. We evaluate the proposed framework by comparing 14 hybrid gradient-free trained RVFL models on a variety of regression and classification datasets. The best-performing models achieve classification accuracies of 96%, 92%, and 85% on the breast cancer, waveform, and magic telescope datasets, respectively. For regression, R-scores of 0.70, 0.89, and 0.80 are observed on the diabetes, Boston housing, and California housing datasets. Additionally, we compare three hybrid RVFL models with GridSearchCV and RandomizedSearchCV on the digits dataset. The results show that our hybrid models yield better performance while requiring significantly less computational time. This suggests that our proposed framework can serve as a critical resource for researchers and practitioners who are seeking practical and resilient approaches to real-world issues. The source code of the library is accessible to the public on the GitHub repository: https://github.com/thieu1995/ GrafoRVFL.

Education is a fundamental sector in all countries, where in some countries students com-pete to get an educational grant due to its high cost. The incorporation of artificial intelli-gence in education holds great promise for the advancement of educational systems and pro-cesses. Educational data mining involves the analysis of data generated within educational environments to extract valuable insights into student performance and other factors that enhance teaching and learning. This paper aims to analyze the factors influencing students' performance and consequently, assist granting organizations in selecting suitable students in the Arab region (Jordan as a use case). The problem was addressed using a rule-based tech-nique to facilitate the utilization and implementation of a decision support system. To this end, three classical rule induction algorithms, namely PART, JRip, and RIDOR, were em-ployed. The data utilized in this study was collected from undergraduate students at the University of Jordan from 2010 to 2020. The constructed models were evaluated based on metrics such as accuracy, recall, precision, and f1-score. The findings indicate that the JRip algorithm outperformed PART and RIDOR in most of the datasets based on f1-score metric. The interpreted decision rules of the best models reveal that both features; the average study years and high school averages play vital roles in deciding which students should receive scholarships. The paper concludes with several suggested implications to support and en-hance the decision-making process of granting agencies in the realm of higher education.

Predicting student performance is a critical aspect of educational systems. Although forecasting a student’s future performance is essential in many applications, it is a challenging process due to various factors. Previous research in this area has mainly focused on comparing machine learning methods to automate student evaluation and predict their final performance. However, there have been limited studies that thoroughly explore the issue of class imbalance using a deep learning approach. Moreover, the large dataset targeting university students makes it well-suited for in-depth analysis and increases the likelihood of obtaining more accurate results. This study presents a deep learning model based on convolution and introduces a comprehensive exploration of oversampling and undersampling methods to address the issue of imbalanced classes. The paper investigates various features and characteristics of undergraduate students at the University of Jordan, utilizing a large dataset collected from the university’s registration unit. These features include demographic information, attributes related to students’ majors, faculties, registrations, courses taken (such as passed, repeated, and completed), as well as their high school averages and performance in the first four semesters. The results demonstrate that the model performs exceptionally well in terms of gmean when predicting students’ excellence. This research project has significant implications and provides valuable insights to the research community and higher education managers, aiding in the development of improved strategies to enhance educational performance. Future researchers can utilize the methods employed in this paper during the data preprocessing stages and implement the demonstrated balancing strategies for further advancements in this field of study.

The global community is awaiting the advent of a self-driving vehicle that is safe, reliable, and capable of navigating a diverse range of road conditions and terrains. This requires a lot of research, study, and optimization. Thus, this work focused on implementing, training, and optimizing a convolutional neural network (CNN) model, aiming to predict the steering angle during driving (one of the main issues). The considered dataset comprises images collected inside a car-driving simulator and further processed for augmentation and removal of unimportant details. In addition, an innovative data-balancing process was previously performed. A CNN model was trained with the dataset, conducting a comparison between several different standard optimizers. Moreover, evolutionary optimization was applied to optimize the model’s weights as well as the optimizers themselves. Several experiments were performed considering different approaches of genetic algorithms (GAs) along with other optimizers from the state of the art. The obtained results demonstrate that the GA is an effective optimization tool for this problem.

The intrusion detection system is responsible for revealing different intrusion activities, including the denial of service, man-in-middle, Mirai, Scan, and other types of intrusion activities. It is used in many applications, including the smart home Internet of Things networks, where security risks threaten the privacy of individuals. In this context, many works were proposed for detecting and classifying the different types of attacks. However, many challenges are identified for this type of problem, such as the large amount of data available, the imbalanced nature of the data, and the quality of detection and classification outcomes. This paper aims to address these challenges by proposing an approach that considers a metaheuristic-based random weight neural network to detect intrusion activities and classify the different types and subtypes of activities. The following points summarize the contribution of this paper. First, the automatic tuning of the neural network parameters where the weights, biases, regularization value, the number of neurons, and the type of activation function are optimized by different metaheuristic algorithms to produce high-quality results. Second, the proposed approach adopts a clustering with reduction technique to tackle the challenge of processing large volumes of data. Third, oversampling the dataset is also embedded in the proposed approach to avoid a biased classification of the majority class. The experiments are conducted based on a large dataset with more than half a million instances. The results show that the proposed approach outperforms the other classification approaches in geometric mean (G-Mean) and has promising results.

Advanced modern technology and the industrial sustainability theme have contributed to the implementation of composite materials for various industrial applications. Bio-composites are among the desired alternatives for green products. However, to properly control the performance of bio-composites, predicting their constituent properties is of paramount importance. This work introduces an innovative, evolving genetic programming tree model for predicting the mechanical properties of natural fibers for the first time based upon several inherent chemical and physical properties. Cellulose, hemicellulose, lignin, and moisture contents, as well as the Microfibrillar angle of various natural fibers, were considered to establish the prediction models. A one-hold-out methodology was applied for the training/testing phases. Robust models were developed utilizing evolving genetic programming tree models to predict the tensile strength, Young’s modulus, and the elongation at break properties of the natural fibers. It was revealed that the Microfibrillar angle was dominant and capable of determining the ultimate tensile strength of the natural fibers by 44.7%, comparable to other considered properties, while the impact of cellulose content in the model was only 35.6%. This would facilitate utilizing artificial intelligence to predict the overall mechanical properties of natural fibers without exhausting experimental efforts and cost to enhance the development of better green composite materials for various industrial applications. Doi: 10.28991/ESJ-2023-07-06-02 Full Text: PDF

This article reports on designing and implementing a multiclass sentiment classification approach to handle the imbalanced class distribution of Arabic documents. The proposed approach, sentiment classification of Arabic documents (SCArD), combines the advantages of a clustering-based undersampling (CBUS) method and an ensemble learning model to aid machine learning (ML) classifiers in building accurate models against highly imbalanced datasets. The CBUS method applies two standard clustering algorithms: K -means and expectation–maximization, to balance the ratio between the major and the minor classes by decreasing the number of the major class instances and maintaining the number of the minor class instances at the cluster level. The merits of the proposed approach are that it does not remove the majority class instances from the dataset nor injects the dataset with artificial minority class instances. The resulting balanced datasets are used to train two ML classifiers, random forest and updateable Naïve Bayes, to develop prediction data models. The best prediction data models are selected based on F1-score rates. We applied two techniques to test SCArD and generate new predictions from the imbalanced test dataset. The first technique uses the best prediction data models. The second technique uses the majority voting ensemble learning model, which combines the best prediction data models to generate the final predictions. The experimental results showed that SCArD is promising and outperformed the other comparative classification models based on the F1-score rates.

Life expectancy prediction helps in enhancing healthcare services for individuals. The complexity of the factors affecting the prediction process, such as lifestyle, smoking, and exercise, makes the prediction results less reliable. Optimizing machine learning tasks using metaheuristic algorithms is very common for several applications. They can be used for parameter tuning of a certain algorithm or feature selection of the dataset. In this study, the life expectancy dataset is experimented with by applying three metaheuristic optimization approaches: the k-nearest neighbor regression algorithm parameter tuning, the feature selection, and a hybrid approach of parameter tuning and feature selection. The results show that the hybrid approach generated the best results with the advancement of 81% and 7.02 for the R-Squared (R 2 ) and the Root Mean Square Error (RMSE), respectively, compared to the K-Nearest Neighbors Regression (KNNR) without any optimization. The best results are obtained by the Grey Wolf Optimizer (GWO) algorithm with an n-neighbors value of 2, a uniform value for the weights parameter, and the Manhattan for the metric parameter. The Schooling, under-five deaths, and Year features are observed to be the most selected features by the metaheuristic algorithm.

Text classification has many applications in various fields; such as news categorization, sentiment analysis, E-mail spam filtering, and others. However, handling textual data is a challenging task owing to the potentially massive number of features (words). The presence of redundant irrelevant features deteriorates the performance of a learning algorithm and makes the process of text classification more complex. This research conducts a comparison study of several filtering-based feature se-lection methods in the context of Arabic text classification. Arabic is a highly complex language syntactically and morphologically which leads to more complicated learning tasks. Proposing a ro-bust classification model is demanding. Remarkably, integrating filtering approaches results in significant improvements in the performance of classification algorithms.

Online reviews are important information that customers seek when deciding to buy products or services. Also, organizations benefit from these reviews as essential feedback for their products or services. Such information required reliability, especially during the Covid-19 pandemic which showed a massive increase in online reviews due to quarantine and sitting at home. Not only the number of reviews was boosted but also the context and preferences during the pandemic. Therefore, spam reviewers reflect on these changes and improve their deception technique. Spam reviews usually consist of misleading, fake, or fraudulent reviews that tend to deceive customers for the purpose of making money or causing harm to other competitors. Hence, this work presents a Weighted Support Vector Machine (WSVM) and Harris Hawks Optimization (HHO) for spam review detection. The HHO works as an algorithm for optimizing hyperparameters and feature weighting. Three different language corpora have been used as datasets, namely English, Spanish, and Arabic in order to solve the multilingual problem in spam reviews. Moreover, pre-trained word embedding (BERT) has been applied alongside three-word representation methods (NGram-3, TFIDF, and One-hot encoding). Four experiments have been conducted, each focused on solving and demonstrating different aspects. In all experiments, the proposed approach showed excellent results compared with other state-of-the-art algorithms. In other words, the WSVM-HHO achieved an accuracy of 88.163%, 71.913%, 89.565%, and 84.270%, for English, Spanish, Arabic, and Multilingual datasets, respectively. Further, a deep analysis has been conducted to investigate the context of reviews before and after the COVID-19 situation. In addition, it has been generated to create a new dataset with statistical features and merge its previous textual features for improving detection performance.

The Covid-19 pandemic has led to an increase in the awareness of and demand for telemedicine services, resulting in a need for automating the process and relying on machine learning (ML) to reduce the operational load. This research proposes a specialty detection classifier based on a machine learning model to automate the process of detecting the correct specialty for each question and routing it to the correct doctor. The study focuses on handling multiclass and highly imbalanced datasets for Arabic medical questions, comparing some oversampling techniques, developing a Deep Neural Network (DNN) model for specialty detection, and exploring the hidden business areas that rely on specialty detection such as customizing and personalizing the consultation flow for different specialties. The proposed module is deployed in both synchronous and asynchronous medical consultations to provide more real-time classification, minimize the doctor effort in addressing the correct specialty, and give the system more flexibility in customizing the medical consultation flow. The evaluation and assessment are based on accuracy, precision, recall, and F1-score. The experimental results suggest that combining multiple techniques, such as SMOTE and reweighing with keyword identification, is necessary to achieve improved performance in detecting rare classes in imbalanced multiclass datasets. By using these techniques, specialty detection models can more accurately detect rare classes in real-world scenarios where imbalanced data is common.

Identifying fraudulent transactions and preventing unauthorized individuals from revealing credit card information are essential tasks for different financial entities. Fraud detection systems are used to apply this task by identifying the fraudulent transactions from the normal ones. Usually, the data used for fraud detection is imbalanced, containing many more instances of normal transactions than fraudulent ones. This causes diminished classification task results because it is hard to train a classifier that distinguishes between them. Another problem is caused by many features under study for the fraud detection task. This paper utilizes different metaheuristic algorithms for feature selection to solve the problem of unneeded features and uses the Synthetic Minority Oversampling TEchnique (SMOTE) to solve the imbalance problem of the data using different classification algorithms. The metaheuristic algorithms include Particle Swarm Optimization (PSO), Salp Swarm Algorithm (SSA), Grey Wolf Optimizer (GWO), and A Multi-Verse Optimizer (MVO), whereas the classification algorithms include Logistic Regression (LR), Decision Tree (DT), and Naive Bayes (NB) algorithms. The results show that applying the oversampling technique generated better results for the G-Mean and Recall values, while the feature selection process enhanced the results of almost all the classification algorithms.

Security of IoT networks is extremely necessary nowadays especially due to the rise of intrusion attacks in IoT networks that are not easily detected and classified. Intrusion detection systems (IDS) focus on both supervised and unsupervised learning techniques to classify intrusion detection using IoT environmental data. Yet most of the time, the huge quantity of data is imbalanced and contains multiclasses. This chapter proposes an approach of multiclass classification using XGBoost and CatBoost to classify the category type of the intrusion attack with the application of the variance threshold feature selection to adopt valid predictions along with the support vector machine and synthetic minority oversampling technique (SVM-SMOTE) to solve imbalanced distribution of the dataset in addition to a way of specifying the highest influential feature variables using the SHAP technique to improve prediction power. The experiments were evaluated based on precision, recall, f1-score, and G-mean where they are then compared against other basic classifiers. The results show that the XGBoost classification method with a variance threshold of 0.8 as well as SVM-SMOTE give better results than other classification techniques. Also, SHAP indicates that the most important features for the prediction of the category labels are Src_Port and Flow_Duration.

[This corrects the article DOI: 10.1371/journal.pone.0290581.].

This paper proposes a fully automated vision-based system for real-time COVID-19 personal protective equipment detection and monitoring. Through this paper, we aim to enhance the capability of on-edge real-time face mask detection as well as improve social distancing monitoring from real-live digital videos. Using deep neural networks, researchers have developed a state-of-the-art object detector called "You Only Look Once Version Five" (YOLO5). On real images of people wearing COVID19 masks collected from Google Dataset Search, YOLOv5s, the smallest variant of the object detection model, is trained and implemented. It was found that the Yolov5s model is capable of extracting rich features from images and detecting the face mask with a high precision of better than 0.88 mAP_0.5. This model is combined with the Density-Based Spatial Clustering of Applications with Noise method in order to detect patterns in the data to monitor social distances between people. The system is programmed in Python and implemented on the NVIDIA Jetson Xavier board. It achieved a speed of more than 12 frames per second. Doi: 10.28991/ESJ-2023-SPER-05 Full Text: PDF

Online media has an increasing presence on the restaurants’ activities through social media websites, coinciding with an increase in customers’ reviews of these restaurants. These reviews become the main source of information for both customers and decision-makers in this field. Any customer who is seeking such places will check their reviews first, which usually affect their final choice. In addition, customers’ experiences can be enhanced by utilizing other customers’ suggestions. Consequently, customers’ reviews can influence the success of restaurant business since it is considered the final judgment of the overall quality of any restaurant. Thus, decision-makers need to analyze their customers’ underlying sentiments in order to meet their expectations and improve the restaurants’ services, in terms of food quality, ambiance, price range, and customer service. The number of reviews available for various products and services has dramatically increased these days and so has the need for automated methods to collect and analyze these reviews. Sentiment Analysis (SA) is a field of machine learning that helps analyze and predict the sentiments underlying these reviews. Usually, SA for customers’ reviews face imbalanced datasets challenge, as the majority of these sentiments fall into supporters or resistors of the product or service. This work proposes a hybrid approach by combining the Support Vector Machine (SVM) algorithm with Particle Swarm Optimization (PSO) and different oversampling techniques to handle the imbalanced data problem. SVM is applied as a machine learning classification technique to predict the sentiments of reviews by optimizing the dataset, which contains different reviews of several restaurants in Jordan. Data were collected from Jeeran, a well-known social network for Arabic reviews. A PSO technique is used to optimize the weights of the features, as well as four different oversampling techniques, namely, the Synthetic Minority Oversampling Technique (SMOTE), SVM-SMOTE, Adaptive Synthetic Sampling (ADASYN) and borderline-SMOTE were examined to produce an optimized dataset and solve the imbalanced problem of the dataset. This study shows that the proposed PSO-SVM approach produces the best results compared to different classification techniques in terms of accuracy, F-measure, G-mean and Area Under the Curve (AUC), for different versions of the datasets.