The digitization of the transportation system has encouraged online ticket booking platforms such as the RedBus application. RedBus helps the public to order tickets from various bus operators in Indonesia. Users can download and view reviews of the RedBus application through the Google Play Store. Reviews are important for both users and developers. However, there are users who give high ratings but have negative reviews. Therefore, it is necessary to analyze user reviews to classify these reviews into positive and negative sentiments class. The total data used is 500 reviews with details of 250 reviews in each positive and negative class. The stages for classifying are manual data labeling, text preprocessing to change the data to be more structured, weighting by the TF-RF and TF-IDF methods, classification by the Naive Bayes algorithm, and testing by the confusion matrix and k-fold cross validation. This study compares 2 word weighting methods, namely TF-RF and TF-IDF to obtain optimal accuracy results. The best accuracy results obtained were using the ratio of training data and test data of 90%:10% with TF-IDF weighting and cross validation testing with a value of k = 10. The average accuracy obtained is 93.56% accuracy, 93.97% precision, 93.57% recall, 94.68% specificity, 93.53% f-measure.
Copyrights © 2023