Sentiment analysis is an essential task in natural language processing, and it involves evaluating attitudes, opinions, and emotions related to different entities. However, most research on sentiment analysis has been focused on English, leaving a gap for more diverse language studies, especially in Arabic. The prevalence of behaviors such as sharing private messages, spreading rumors, and making sexual comments has drawn significant attention due to their severe negative social consequences. Detecting bullying texts or messages on social media has become a fundamental interest for researchers worldwide. This study focuses on creating an efficient online method to identify abusive and bullying messages by combining natural language processing (NLP) with machine learning. The research assesses the performance of different machine learning algorithms using frequency-inverse document frequency (TFI_DF). This chapter delves into significant research on sentiment analysis of Arabic text using deep learning algorithms. It offers a detailed examination of Sentiment analysis to detect cyberbullying, outlining the methodologies employed for data collection, preprocessing, and analysis. An extensive literature review identified research gaps and practical techniques for cyberbullying detection across various languages, highlighting significant potential for advancement in Arabic. Consequently, this research evaluates specific machine learning algorithms for classifying Arabic datasets obtained from Twitter (now known as X). The Machine learning algorithms under investigation include Support Vector Machine (SVM), Support Vector Classifier (SVC), Naïve Bayes (NB), Multinominal Naïve Bayes (MNB), Random Forest (RF), Logistic Regression (LR), Gradient Boosting (GBoost), K-nearest Neighbor (KNN), Adaptive Boosting (AdaBoost), Decision Tree (DT) and eXtreme Gradient Boosting (XGBoost), selected for their proven effectiveness in similar contexts. These algorithms’ performance was measured using well-established metrics such as accuracy, precision, recall, and F1-score. RF emerged as the top performer with an accuracy of 80.9%, showing promise compared to current state-of-the-art methods. This research has the potential to significantly advance cyberbullying detection and stimulate further studies in this field.