The exponential increase in data has created opportunities and challenges for machine learning, particularly in annotating unlabeled datasets, a task that is time-intensive and often requires domain expertise. Clustering algorithms, such as k-means, offer an alternative approach but face limitations, including susceptibility to local minima. Metaheuristic algorithms like Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) have been applied to enhance clustering quality. This study extends prior work by comparing the Salp Swarm Algorithm (SSA), Harris Hawks Optimization (HHO), Heap-Based Optimizer (HBO), and Golden Eagle Optimizer (GEO) for clustering problems, focusing on three population initialization strategies: random initialization, K-means initialization, and K-means++ initialization. While random initialization generates clusters randomly, K-means and K-means++ methods integrate centroid spread techniques to enhance optimization. A diverse range of datasets was used, from small, well-known benchmarks to larger, higher-dimensional datasets to evaluate the performance of these approaches against GA, PSO, Differential Evolution (DE), and Artificial Bee Colony (ABC). The results reveal that hybrid methods combining K-means initialization with metaheuristic optimization significantly reduce within-cluster distances, outperforming traditional clustering methods. These findings suggest improved cluster quality, offering potential benefits for downstream machine learning tasks.

Ontologies offer a powerful means of structuring medical knowledge for AI-driven diagnosis. This paper investigates the conceptual design of a framework that integrates ontologies and logical reasoning to automate the diagnosis of nutrition-related medical conditions. There is a significant gap in the literature on direct diagnosis using ontologies and logical reasoning for such conditions. By proposing a theoretical framework, this work aims to enhance diagnostic accuracy, efficiency, and reliability while providing explainable diagnoses. It is important to note that this study is part of an ongoing Ph.D. thesis and that the implementation of the proposed framework is planned in future work. This approach has the potential to improve patient outcomes and reduce healthcare costs significantly.

Automated code generation from natural language is a growing challenge in software engineering, enabling the direct translation of requirements into executable code. Machine learning-based automatic code synthesis emerged as a key solution, with transformer-based models playing a pivotal role in enhancing this process. This paper addresses the challenges of generating accurate, functional code efficiently. This paper presents a transformer-based T5 model fine-tuned on the multilingual XLCoST dataset for code generation. The model is evaluated using a novel systematic multi-level evaluation framework. The framework evaluates the model’s across three distinct task complexity levels ranging from beginner to advanced. The evaluation results highlight the model’s versatility and robustness. Furthermore, the model achieves a BLEU score of 30.093, demonstrating competitive performance compared to open-source baselines. The model’s code generation abilities accelerates software development by reducing manual coding efforts, thereby improving workflow efficiency and software quality. These results confirm that transformer-based models are effective tools for automating software development, with promising implications for the future.

Stylometric features embedding is crucial for author attribution, using statistical and linguistic patterns to identify unique stylistic traits of writers. This method focuses on word length, sentence complexity, and punctuation for comprehensive authorship profiles. Analyzing these features reveals patterns in individual writing styles, enhancing attribution accuracy. This study evaluates various classification algorithms and feature extraction techniques to distinguish texts by the same author from those by others. It faces challenges due to nuanced stylistic differences, text length inconsistencies, genre diversity, and contextual variations. The study employs TFIDF for feature extraction, assessing word importance in documents. Results show TFIDF effectively captures authorship characteristics, with non-linear Support Vector Machines performing strongly across feature dimensions. Optimizing SVM parameters through grid search improves performance with larger feature sizes, achieving peak accuracy of 81.13

The Internet of Things (IoT) has evolved as a significant area of impact and opportunity due to the billions of connected devices that have been distributed around the world. IoT devices, on the other hand, are susceptible to being compromised and hacked. When it comes to computing power and storage capacity, these Internet of Things devices are more vulnerable to cyberattacks than traditional endpoints such as smartphones, tablets, and laptops. This study introduces and assesses a machine learning-based cyberattack detection system. The suggested method employs a Support Vector Machine (SVM) classifier with the Harris Hawks Optimization (HHO) algorithm. The HHO technique improves SVM classifier hyperparameters, while the SVM performs malicious and normal stream classification based on the best-chosen model, and produces the optimal solution for feature weighting. The utility and capacity of the suggested approach to improve detection performance is demonstrated through proper scientific testing utilizing the LITNET-2020 benchmark dataset against six well-known classification algorithms and four metaheuristic-based classifiers using five reliable assessment measures.

Over the past decade, knowledge resources have become central to advancing digital health, particularly in nutrition research. Nutrition is crucial for human growth, development, metabolism, and immunity, making it a key focus in public health and personalized medicine. Ontologies and Knowledge Graphs (KGs) have emerged as essential tools for addressing challenges in this field. This systematic review, conducted following the PRISMA guidelines, focuses primarily on the use of ontologies in human nutrition, including their applications, development methodologies, construction resources, evaluation criteria, and associated challenges. KGs are considered in a secondary role to support the preliminary analysis and provide insights into related applications. The review covers studies published between 2014 and 2024. A comprehensive search across Scopus, Web of Science, IEEE Xplore, and ACM Digital Library identified 58 studies, including 45 utilizing ontologies and 13 employing KGs. The findings reveal a growing use of these tools to develop personalized recipes, nutrient intake recommendations, food suggestions, dietary plans, and recommendation systems. However, significant gaps persist in the standardization of methodologies, resource integration, and evaluation criteria, limiting interoperability and scalability for personalized applications. Ontology-based systems dominate knowledge representation, while KGs are being explored for personalization in diet planning and data integration. To our knowledge, this is the first systematic review to exclusively focus on the intersection of ontologies with a preliminary analysis of KGs in human nutrition. By bridging computational techniques with nutritional science, this review establishes the foundation for innovative applications in digital health and personalized nutrition, contributing to improved public health outcomes.

Semantic Interoperability (SI) enables cross-domain data integration by allowing diverse systems to share and process information effectively. While existing reviews focus on general AI-driven interoperability, this systematic literature review (SLR) is the first to exclusively analyze the integration of Large Language Models (LLMs) with SI. This SLR uniquely evaluates LLMs' role in schema alignment, knowledge integration, and security risks. It also introduces a novel taxonomy and identifies challenges like bias propagation and computational costs, providing a new research framework for adversarial robustness, ethical AI, and real-world SI optimization.This is an open access article under the CC BY-NCND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

The rapid growth of data has posed significant challenges for machine learning, particularly in the context of unlabeled datasets. While clustering techniques offer potential solutions, the k-means algorithm often suffers from local minima. This study introduces hybrid approaches combining the Salp Swarm Algorithm (SSA) and Harris Hawks Optimization (HHO) with different initialization strategies (random, K-means, K-means++) to enhance clustering performance. Experimental results on ten UCI benchmark datasets demonstrate that these hybrid approaches significantly outperform traditional K-means and Kmeans++ methods in minimizing within-cluster distances. This suggests improved cluster quality, which can be valuable for various downstream applications.

Smart cities represent the future of urban evolution, characterized by the intricate integration of the Internet of Things (IoT). This integration sees everything, from traffic management to waste disposal, governed by interconnected and digitally managed systems. As fascinating as the promise of such cities is, they have its challenges. A significant concern in this digitally connected realm is the introduction of fake clients. These entities, masquerading as legitimate system components, can execute a range of cyber-attacks. This research focuses on the issue of fake clients by devising a detailed simulated smart city model utilizing the Netsim program. Within this simulated environment, multiple sectors collaborate with numerous clients to optimize performance, comfort, and energy conservation. Fake clients, who appear genuine but with malicious intentions, are introduced into this simulation to replicate the real-world challenge. After the simulation is configured, the data flows are captured using Wireshark and saved as a CSV file, differentiating between the real and fake clients. We applied MATLAB machine learning techniques to the captured data set to address the threat these fake clients posed. Various machine learning algorithms were tested, and the k-nearest neighbors (KNN) classifier showed a remarkable detection accuracy of 98 77%. Specifically, our method increased detection accuracy by 4.66%, from 94.02% to 98.68% over three experiments conducted, and enhanced the Area Under the Curve (AUC) by 0.49%, reaching 99.81%. Precision and recall also saw substantial gains, with precision improving by 9.09%, from 88.77% to 97.86%, and recall improving by 9.87%, from 89.23% to 99.10%. The comprehensive analysis underscores the role of preprocessing in enhancing the overall performance, highlighting its superior performance in detecting fake IoT clients in smart city environments compared to conventional approaches. Our research introduces a powerful model for protecting smart cities, merging sophisticated detection techniques with robust defenses.

Introduction As many web platforms adopt collaborative content editing models, the gender gap is addressed as one of the chief concerns in using technology to restrict content editing by one gender. Objective This study aims to analyze the Arabic Wikipedia, the largest collaborative content editing platform on the Arabic web, in terms of gender behavior and differences in user activities. Methods This study is the first to address the gender gap in Arabic Wikipedia, characterize users’ gender through their behavior, and then address changes in characteristics over the past five years. This study analyzes parts of Arabic Wikipedia offline by linking article pages and page edit histories to user profiles of known genders. Results This study reported that a gender gap exists in Arabic Wikipedia. The results reported differences over the past five years between both genders in terms of tasks and user behavior. One aspect that indicated similarity is the period of active time over months/years. Differences were observed in the reported number of increasing users, activities, responsibilities, and average actions performed. Conclusion The results reveal a vast gap in terms of gender behavior in Wikipedia activities. Moreover, the results reveal that some administrative activities are disclosed to men more than to women.

The development of emotion detection technology has emerged as an efficient possibility in the corporate sector due to the nearly limitless uses of this new discipline, particularly with the unceasing propagation of social data. In recent years, the electronic marketplace has witnessed the establishment of various start-up businesses with an almost sole focus on building new commercial and open-source tools and APIs for emotion detection and recognition. Yet, these tools and APIs must be continuously reviewed and evaluated, and their performances should be reported and discussed. There is a lack of research to empirically compare current emotion detection technologies in terms of the results obtained from each model using the same textual dataset. Also, there is a lack of comparative studies that apply benchmark comparisons to social data. This study compares eight technologies: IBM Watson Natural Language Understanding, ParallelDots, Symanto - Ekman, Crystalfeel, Text to Emotion, Senpy, Textprobe, and Natural Language Processing Cloud. The comparison was undertaken using two different datasets. The emotions from the chosen datasets were then derived using the incorporated APIs. The performance of these APIs was assessed using the aggregated scores they delivered and the theoretically proven evaluation metrics such as the micro-average of accuracy, classification error, precision, recall, and f1-score. Lastly, the assessment of these APIs incorporating the evaluation measures is reported and discussed.

In this study, an Aspect-based Sentiment Analysis (ABSA) model was developed to classify restaurants' reviews in the Arabic language based on four predefined aspects: price, cleanliness, food quality, and service. A hybrid approach that combines machine learning with domain-specific dictionaries and sentiment word lists was proposed for ABSA. More than 3,000 reviews were collected from a restaurant reviews website. The reviews were annotated using a crowdsourcing method. The annotated reviews were pre-processed, then the dictionaries and sentiment word lists were extracted from the dataset. Moreover, a filter-based feature selection approach using the Chi 2 method was applied to reduce the number of representative features. Four aspect models were built using Support Vector Machine (SVM) and another four models were built using Naïve Bayes (NB) classifiers, one model for each aspect. The models were evaluated using Accuracy, Precision, Recall, and F-Measure. The results were promising, as the price aspect model achieved the highest results by applying the SVM classifier with Accuracy 84.47%, Precision 84.3%, Recall 84.5%, and F-Measure 84.3%.

An enormous number of microblogs are being created and posted on the web each day. Many of these microblogs are repetitive in terms of content and similar in terms of topic. Being able to detect repetitive content can support various applications such as question answering and trendy topic detection. In this research, we aim to propose a model to detect paraphrasing among Arabic tweets, in addition to identifying tweets belonging to the same topic. The proposed model is based on Latent Dirichlet Allocation (LDA) topic modeling, as well as, semantic text expansion utilizing external resources i.e. BabelNet and Wikipedia. Tweets from multiple Arabic news agencies were collected, preprocessed, and divided into two groups. The first group was used to build the topic modeling and the other group of tweets was paired and classified based on the topic distributions. The results are promising in terms of precision on tweet pairs with a certain time overlap. The best-reported precision is 80.1% achieved using Wikipedia embedded content on the stemmed text mode with a large number of LDA topics.

Some classic machine translation (MT) Evaluation methods, such as the bilingual evaluation understudy score (BLEU), have notably underperformed in evaluating machine translations for morphologically rich languages like Arabic. However, the recent remarkable advancements in the domain of word vectors and sentence vectors have opened up new research avenues for low-resource languages. This paper proposes a novel linguistic-based evaluation method for English-translated sentences in Arabic. The proposed approach includes penalties based on length, positions, and context-based schemes such as part-of-speech tagging (POS) and multilingual sentenceBERT (SBERT) models for machine translation evaluation. The proposed technique is tested using pearson correlation as a performance evaluation parameter and compared with state-of-the-art techniques. The experimental results demonstrate that the proposed model evidently outperforms other MT evaluation methods such as BLEU.

The release of millions of financial documents, which has been known as the ‘WikiLeaks’ of the financial world (a.k.a. ‘Panama Papers’), has dragged global attention in how highly structured means applied by some of the elite to conceal their financial assets. Consequently, significant financial corruption allegations were raised. We concentrate on a somewhat overlooked region, the Middle East and North Africa (MENA) region. This study aims to use social network analytics to study the information contained in these documents. We are checking the major players in the MENA’s trends and patterns to determine if it matches the known economic powers. The analysis reveals that while the constructed network enjoys some typical characteristics, many interesting observations and properties are worth discussing. Specifically, using the extracted network consisting of 62 987 nodes and 84 692 edges, our social network analysis finding shows that, perhaps surprisingly, the nodes or the social network are not necessarily directly correlated with perceived economic influence.

Context: Predicting the priority of bug reports is an important activity in software maintenance.Bug priority refers to the order in which a bug or defect should be resolved.A huge number of bug reports are submitted every day.Manual filtering of bug reports and assigning priority to each report is a heavy process, which requires time, resources, and expertise.In many cases mistakes happen when priority is assigned manually, which prevents the developers from finishing their tasks, fixing bugs, and improve the quality.Objective: Bugs are widespread and there is a noticeable increase in the number of bug reports that are submitted by the users and teams' members with the presence of limited resources, which raises the fact that there is a need for a model that focuses on detecting the priority of bug reports, and allows developers to find the highest priority bug reports.This paper presents a model that focuses on predicting and assigning a priority level (high or low) for each bug report.Method: This model considers a set of factors (indicators) such as component name, summary, assignee, and reporter that possibly affect the priority level of a bug report.The factors are extracted as features from a dataset built using bug reports that are taken from closed-source projects stored in the JIRA bug tracking system, which are used then to train and test the framework.Also, this work presents a tool that helps developers to assign a priority level for the bug report automatically and based on the LSTM's model prediction.Results: Our experiments consisted of applying a 5-layer deep learning RNN-LSTM neural network and comparing the results with Support Vector Machine (SVM) and K-nearest neighbors (KNN) to predict the priority of bug reports.The performance of the proposed RNN-LSTM model has been analyzed over the JIRA dataset with more than 2000 bug reports.The proposed model has been found 90% accurate in comparison with KNN (74%) and SVM (87%).On average, RNN-LSTM improves the F -measure by 3% compared to SVM and 15.2% compared to KNN.Conclusion: It concluded that LSTM predicts and assigns the priority of the bug more accurately and effectively than the other ML algorithms (KNN and SVM).LSTM significantly improves the average F -measure in comparison to the other classifiers.The study showed that LSTM reported the best performance results based on all performance measures (Accuracy = 0.908, AUC = 0.95, F -measure = 0.892).

Explainable recommendation systems have gained much attention in the last few years. Most of them use textual reviews to provide users with interpretability about why services or products are liked by users or recommended for them. Sentiment analysis has potential advantages to determine the attitudes of users in online communities using websites such as Twitter, Facebook, and YouTube. However, sentiment analysis of textual reviews in explainable recommendation systems seems to be a really challenging task. In this paper, we present a deep learning-based architecture for sentiment analysis to automatically predict the sentiment of reviews, which are considered as explanations of recommendations. It consists of two instances of the prediction model, one with the Long Short-Term Memory (LSTM) method and the other with the Gated Recurrent Unit (GRU) method. We evaluate their performance on one real-world dataset from Amazon and compare them with one state-of-the-art method. The experimental results show that our methods perform better than the baseline approach.

Click fraud is a serious problem facing online advertising business. The malicious intent of clicking online ads either committed by humans or by non-humans, forced financial losses on advertisers utilizing pay-per-click advertising. Non-human traffic is usually designed to inflate web traffic for fraudulent purposes. In this paper, we demonstrate a hybrid approach consisting of two-level fingerprint applied in two phases to detect illegitimate non-human traffic. The first-level fingerprint is a pattern generated using immutable information about a user navigating a website's pages. It will be used in the first traffic illegitimacy detection phase to infer rules about illegitimate non-human traffic from a developed ontology about web traffic legitimacy. The second-level fingerprint is generated using behavioral ad click patterns, which will be used in the second detection phase by applying a Machine-Learning (ML) algorithm. To test the proposed approach, a real commercial website for ads, called Waseet.com, was used. The access logs of the website server were utilized for the purpose of this research. The experiments show that our proposed hybrid approach using the ontology of web traffic illegitimacy and the ML k-NN classifier detects around (98.6%) of fake clicks.

Social media platforms changed from being socialization platforms to serve businesses through advertisements. This research aims at investigating active young users' experience with social media ads by studying the personalization and the usefulness of the ads, and the role of the host architecture of the used platform. The results prove that users' experience was affected by the designated variables: personalization, perceived usefulness, and the host architecture. Specifically, It was found that social media users find social media ads useful, and personalized, and that the perceived usefulness and personalization significantly affect the usage of host architecture which significantly affects users' experience. Additionally, a significant difference is found between clusters of student answers in terms of personalization and perceived usefulness effect on user experience.

Recent advances in information filtering have resulted in effective recommender systems that are able to provide online personalized recommendations to millions of users from all over the world. However, most of these systems ignore the explanation purpose while producing recommendations with high-quality results. Moreover, the classification of reviews given to users as explanations is not fully exploited in previous studies. In this paper, we develop a convolutional neural network-based reviews classification method for explainable recommendation systems. The convolutional neural network is used to extract the reviews features for predicting whether the reviews provided as explanations are positive or negative. Based on such additional information, users can understand not only why certain items are recommended for them but also get support to know the nature of such explanations. We conduct experiments on a dataset from Amazon. The experimental results show that our method outperforms state-of-the-art methods.