GRADIENT BOOSTING METODI VOSITASIDA O‘ZBEK TILI MATNLARINI SENTIMENT TAHLIL QILISH

This article examines the effectiveness and possibilities of sentiment analysis of Uzbek texts using the Gradient Boosting (GB) method. The texts in the corpus were prepared in advance through the stages of cleaning, tokenization, and normalization. The accuracy of the model, created on the basis of the gradient boosting algorithm, was assessed through experimental tests based on such metrics as precision, recall, and F1-score. Due to GB's ability to correct sequential errors, complex phrases in the Uzbek language (for example, "not good," "no bad") were clearly interpreted. Methods such as morphological normalization (separation of word roots) and data balancing increased model reliability by 4%. The research results confirm that they can be used in customer feedback analysis, social media monitoring, and automatic evaluation systems in language services.