Sentiment analysis, a crucial domain within Natural Language Processing (NLP), aims to extract and interpret opinions, emotions, and sentiments from text, offering valuable insights into public sentiment and opinion trends. This study specifically addresses the sentiment analysis of company reviews written in Arabic, focusing on classifying sentiments into positive and negative categories. Due to the Arabic language's complex morphology and extensive vocabulary, specialized techniques are required. To address these challenges, we employed various machine-learning models, including decision tree, Naïve Bayes, logistic regression, and random forest, as well as hybrid approaches that combine these models to enhance accuracy. Our methodology included comprehensive data preprocessing, normalizing characters, eliminating stopwords, tokenizing, and stemming to prepare the data effectively for analysis. The performance of the models was assessed using metrics such as accuracy, precision, recall and F1 score, with results illustrated through detailed charts and tables. Our findings indicate that hybrid models and the random forest algorithm were especially effective in handling the linguistic intricacies of Arabic. This research provides valuable insights for improving sentiment analysis techniques in Arabic, highlighting the potential of customized machine learning approaches for better sentiment classification in morphologically rich languages.