Customer reviews significantly influence purchasing decisions and business strategies, but manual analysis becomes impractical with increasing review volumes. This study presents an AI model for automated customer review analysis using rule-based techniques and machine learning approaches. This study apply Natural Language Processing (NLP) methods including Bag of Words, Hash Vectorization, Word2Vec, and TF-IDF to the YelpZip dataset, utilizing XGBoost, Logistic Regression, Random Forest, and Decision Trees for classification. The RoBERTa transformer enhances sentiment analysis capabilities. The hybrid approach combines rule-based interpretability with machine learning generalizability through weighted aggregation mechanisms. Experimental results demonstrate improved classification performance with 91.8% recall and 86.8% F1-score, providing scalable insights for automated review analysis and fake review detection.