This study explores the efficacy of sentiment analysis in e-commerce by employing a combination of a rule-based classifier and traditional machine learning models. The objective is to automate the categorization of customer feedback into distinct sentiment labels: Positive, Negative, Suggestion, Neutral, and Unknown. Data from 800 feedback sentences were preprocessed and manually labeled to train and validate our models. The rule-based classifier was optimized through comprehensive feature engineering and dictionary refinement. We employed 10-fold cross-validation to fine-tune the classifier’s parameters, ensuring robust performance across different subsets of the data. This optimization process involved iterative improvements to the sentiment word dictionaries and the implementation of heuristic rules to enhance classification accuracy. For comparative purposes, several machine learning models, including Random Forest, Gradient Boosting, Support Vector Machine (SVM), Naive Bayes, Logistic Regression, and K-Nearest Neighbors, were trained using TF-IDF features extracted from the text data. Our findings indicate that the optimized rule-based classifier outperforms traditional machine learning models, achieving an accuracy of 90%, a recall of 90%, an F1 score of 90%, and a precision of 91%. These results demonstrate that the proposed rule-based approach not only accelerates the sentiment analysis process but also provides more reliable outcomes compared to manual methods. This study highlights the potential of automated sentiment analysis systems in enhancing customer satisfaction by delivering timely and accurate insights into customer opinions.