Messaoud Mezati

Social media platforms generate vast streams of emotionally rich textual data, offering valuable opportunities for critical applications, including mental health assessment and the analysis of collective public sentiment. However, detecting emotions in noisy and multilingual content remains challenging, particularly for under-resourced varieties such as dialects. Moreover, supervised learning techniques strongly depend on the availability of manually annotated corpora, whose creation requires substantial human effort and domain expertise. In contrast, unsupervised methods, while avoiding the need for human intervention, often lack sufficient robustness when confronted with the variability and complexity of natural language across diverse linguistic and cultural contexts. We present an ensemble clustering framework that automatically generates emotion labels from Twitter data, without human intervention. Our approach incorporates three emoji-handling strategies in the preprocessing step, enabling diverse semantic representations of emojis. We applied the BERT embeddings combined with PCA for dimensionality reduction within the same experimental framework. An ensemble clustering strategy integrating K-Means, Agglomerative clustering, and Gaussian Mixture Models (GMM) is adopted using multiple ensemble configurations. Experimental evaluation conducted on 10017 English tweets and 4134 Arabic tweets demonstrates that the proposed method achieves a silhouette score of 0.808 on English data using K-Means with Agglomerative and K-Means with GMM ensemble configurations. For Arabic data, silhouette scores of 0.728 and 0.718 are obtained using English and Arabic keywords, respectively. Emoji semantic integration enhances ensemble clustering performance, suggesting its importance for contextual disambiguation. The proposed framework provides a scalable solution for emotion detection in low-resource languages, enabling language-aware applications in multilingual contexts, particularly within linguistically diverse and multilingual populations

Machine learning (ML) in big data frameworks plays a critical role in real-time analytics, decision making, and predictive modeling. Among the most prominent ML libraries for large-scale data processing are Flink-ML, the machine learning extension of Apache Flink, and MLlib, the machine learning library of Apache Spark. This paper provides a comparative analysis of these two frameworks, evaluating their performance, scalability, streaming capabilities, iterative computation efficiency, and ease of integration with external deep learning frameworks. Flink-ML is designed for real-time, event-driven ML applications and provides native support for streaming-based model training and inference. In contrast, Spark MLlib is optimized for batch processing and micro-batch streaming, making it more suitable for traditional machine learning workflows. Experimental results show that training time is nearly identical for both frameworks, with Spark MLlib requiring 4006.4 seconds and Flink-ML 4003.2 seconds, demonstrating comparable efficiency in batch training and streaming-based model updates. Accuracy results show that Flink-ML (74.9%) slightly outperforms Spark MLlib (74.7%), suggesting that continuous learning in Flink-ML may contribute to better generalization. Inference throughput is slightly higher for Spark MLlib (8.4 images/sec) compared to Flink-ML (8.2 images/sec), suggesting that Spark's batch execution provides a slight advantage in processing efficiency. Both frameworks consume the same amount of memory (30.2%), confirming that TensorFlow's deep learning operations dominate resource consumption rather than architectural differences between Spark and Flink. These results highlight the tradeoffs between Flink-ML and Spark MLlib, and guide data scientists and engineers in selecting the appropriate framework based on specific ML workflow requirements and scalability considerations.

Social media has become a vital element of everyday life, shaping domains like business, politics, and personal interactions. Emotions play a critical role in these areas, necessitating accurate detection and interpretation, especially on platforms like Twitter (X), which feature short texts, various data formats (such as words, Emojis, and numbers), and multilingual content, including dialects. This study explores the importance of Emojis and keywords in positively interpreting emotions on Twitter (X). It uses ensemble-clustering techniques, combining different clustering algorithms like KMeans with various methods for a detailed analysis of emotional subtleties in social media discourse. By merging the semantic meanings of Emojis and keywords, a novel clustering ensemble algorithm is proposed to improve emotion detection accuracy. The approach is tested on two datasets: English and Arabic dataset, using the Ekman model, which classifies emotions into six basic categories (joy, sadness, anger, disgust, surprise, and fear). The findings from this integrated method show greater accuracy and precision compared to individual methods, providing valuable insights into public sentiments, enhancing customer satisfaction analysis, and improving social media monitoring tools.

Age detection from children's drawings is an innovative approach to understanding developmental milestones through visual analysis. Traditional methods for determining a child's age often rely on linguistic or cognitive assessments, but visual data such as drawings offer untapped potential for non-invasive analysis. This study explores the use of clustering algorithms to detect age patterns in children's drawings, providing a novel method for age estimation. A dataset of children's drawings, representing different age groups, was collected, and key visual features, such as line thickness, and object proportions, were extracted. These features were analyzed using unsupervised clustering algorithms, including K-means, Agglomerative Clustering, Mean-shift, and others, to group the drawings based on age-related visual characteristics. Among the algorithms tested, Mean-shift achieved the highest performance, with a silhouette score of 0.67 in mapping clusters to the correct age labels. K-means and Agglomerative Clustering exhibited moderate performance, with silhouette scores of 0.57 and 0.46. In contrast, Spectral Clustering and OPTICS performed poorly, with negative silhouette scores, reflecting poorly defined cluster boundaries. Our study demonstrates the potential of clustering algorithms for automatic age detection from visual features in children's drawings, despite challenges such as overlapping features between adjacent age groups. The findings suggest directions for future research, including more complex models that integrate visual and cognitive indicators for enhanced age and developmental assessment. This approach has significant implications for educational psychology, child development, and artificial intelligence.

Artificial Intelligence (AI) has revolutionized various fields, including psychology and behavioral sciences, by offering novel approaches to understanding human development and behavior. Recent advancements in machine learning have enabled the analysis of children's drawings to gain insights into their cognitive and emotional states. Despite progress, there remains a significant challenge in accurately detecting the age of older children from their drawings. While numerous studies have focused on early childhood development through drawings, there is a paucity of research on late age detection using AI. Specifically, the ability to determine the age of older children from their artwork remains underexplored and imprecise. Our research addresses this gap by developing an AI-based model that accurately detects the age of older children based on their drawings. We employ a deep learning approach, utilizing Long short-term memory (LSTM) networks to analyze various features in children's artwork that correlate with age. We compiled a dataset of drawings from children aged 2-10 years. Our model was trained and validated on this dataset, achieving a high accuracy rate in age detection. Key features such as complexity, use of space, and thematic content were analyzed. The results demonstrate that our AI model can predict the age of children with a significant degree of accuracy, outperforming existing methods. These findings enhance our understanding of cognitive and artistic development in older children. The implications of this research are far-reaching, offering potential applications in educational psychology, developmental assessments, and personalized education plans.

Radiation oncology is a specialized medical field that employs various forms of radiation therapy to treat cancer and other diseases. Decision-making in this domain often involves intricate considerations. This paper investigates applying natural language processing (NLP) techniques for classifying clinical text data to recommend the appropriate radiation therapy modality proton therapy or photon therapy. We use the Clinical Text Classification subset of the ROND database, which consists of only 100 manually annotated cancer cases. We start by pre-processing the text data, extracting TF-IDF features, and applying random oversampling to handle the class imbalance issue. We train and evaluate several machine learning models, including support vector machines (SVM), random forest, decision trees, and naive Bayes on this textual data alone. The SVM model achieved the highest accuracy of 95% after oversampling. To enhance interpretability, we apply SHAP values method which provide insights into which features are most important for the model's predictions towards recommending therapy and how they impact those predictions, which be crucial for understanding and interpreting machine learning model.