Noreddine Gherabi

Depression profoundly affects emotional states, behaviors, and overall quality of life. In the digital era, patients increasingly share their experiences through online comments, offering valuable yet complex emotional narratives. However, traditional sentiment analysis methods, which often assign a single emotional label to an entire text, fail to capture the nuanced intra-comment emotional fluctuations, overlooking the evolving and sometimes contradictory nature of emotional expression. This study introduces a novel sentiment analysis framework that models intra-comment emotional flow as a Brownian-like trajectory. Inspired by the random motion of particles in physics, the proposed method treats each sentence as a time step and assigns it a sentiment polarity score. The cumulative sequence of these scores constructs an emotional curve that reflects the dynamic affective progression throughout the comment, effectively simulating a Brownian path. From these trajectories, quantitative indicators—including drift, variance, -signal changes, trajectory length ( ), maximum emotional level ( ), minimum emotional level ( ), and amplitude — are extracted, providing a detailed and structured understanding of emotional instability. To uncover patterns across comments, the K-Means clustering algorithm is applied, enabling the automatic grouping of comments with similar emotional dynamics. In addition, Principal Component Analysis (PCA) is used to reduce dimensionality and facilitate the visualization and interpretation of emotional profiles. The proposed approach is applied to a corpus of comments from depressed patients to identify characteristic emotional patterns and examine differences according to the type of treatment. This work highlights the relevance of stochastic modeling for advancing sentiment analysis in complex psychological contexts and offers a tool to support remote monitoring, personalized interventions, and early detection of emotional relapse.

Many traffic-sign detection systems are available to assist drivers with particular conditions such as small and distant signs, multiple signs on the road, objects similar to signs, and other challenging conditions. Real-time object detection is an indispensable aspect of these detection systems, with detection speed and efficiency being critical parameters. In terms of these parameters, to enhance performance in road-sign detection under diverse conditions, we proposed a comprehensive methodology, SSAM_YOLOv5, to handle feature extraction and small-road-sign detection performance. The method was based on a modified version of YOLOv5s. First, we introduced attention modules into the backbone to focus on the region of interest within video frames; secondly, we replaced the activation function with the SwishT_C activation function to enhance feature extraction and achieve a balance between inference, precision, and mean average precision (mAP@50) rates. Compared to the YOLOv5 baseline, the proposed improvements achieved remarkable increases of 1.4% and 1.9% in mAP@50 on the Tiny LISA and GTSDB datasets, respectively, confirming their effectiveness.

Sentiment analysis (SA) plays a central role in understanding the multidimensional nature of comments on social media platforms like YouTube, Twitter, Facebook, and health forums. This article presents a multi-aspect, multi-sentiment annotation of chronic disease-related comments, highlighting the importance of connecting aspects, such as disease category and treatment, with corresponding sentiments (positive, negative, and neutral). The integration of ontologies enhances semantic consistency, enabling a comprehensive understanding of user experiences related to chronic diseases. The purpose of this paper is to bridge the gap between unstructured patient-generated content and actionable insights for healthcare professionals. A structured approach employing two ontologies—one for chronic disease aspects and another for sentiment classification—enables the linking of disease-related comments with associated emotions, supporting comprehensive SA. A multi-label classification model is trained to simultaneously predict multiple aspects and sentiments within a single comment, addressing the native complexities of sentiment expression. The article concludes with an evaluation of the model’s predictions against real-world annotated data to assess its effectiveness. Our approach achieves promising results, demonstrating its ability to accurately link comments to their corresponding aspects and sentiments.

The increasing complexity and sophistication of cyber threats underscore the critical need for advanced threat detection mechanisms within Security Operations Centers (SOCs) to effectively mitigate risks and enhance cybersecurity resilience. This study enhances the capabilities of Wazuh, an open-source Security Information and Event Management (SIEM) system, by addressing its primary limitation: high false-positive rates in rule-based detection. We propose a hybrid approach that integrates machine learning (ML) techniques—specifically, Random Forest (RF) and DBSCAN—into Wazuh’s detection pipeline to improve both accuracy and operational efficiency. Experimental results show that RF achieves 97.2% accuracy, while DBSCAN yields 91.06% accuracy with a false-positive rate of 0.0821, significantly improving alert quality. Real-time deployment requirements are rigorously evaluated, with all models maintaining end-to-end processing latencies below 100 milliseconds and 95% of events processed within 500 milliseconds. Scalability testing confirms linear performance up to 500 events per second, with an average processing latency of 45 milliseconds under typical SOC workloads. This integration demonstrates a practical, resource-efficient solution for enhancing real-time threat detection in modern cybersecurity environments.

The rapid rise of information sharing on social media has amplified the spread of fake news, making its detection increasingly critical. As fake news continues to proliferate, the need for efficient detection mechanisms has become more urgent to protect users from misinformation and disinformation. This paper presents a comparative analysis of multiple machine learning models for detecting text based fake news on social media platforms. Using models such as gradient boosting, XGBoost, and linear support vector classifier (SVC) on the Infor mation Security and Object Technology (ISOT) fake news dataset, the study demonstrates that gradient boosting achieves the highest accuracy of 99.61%, while XGBoost provides a strong balance with 99.59% accuracy and a signifi cantly lower execution time, making it more suitable for real-time applications. These results offer valuable insights into the trade-offs between accuracy and computational efficiency, contributing to the development of more practical de tection systems and future research in the field.