Analisis Sentimen Pengguna Aplikasi Taobao Berdasarkan Rating dan Ulasan di Google Play Store Menggunakan Algoritma Naive Bayes

Abstract

This study performs sentiment analysis on Taobao user reviews from the Google Play Store using the Naive Bayes algorithm. A total of 500 reviews were collected through web scraping and preprocessed with case folding, tokenizing, stopword removal, and stemming. Sentiment labels (positive, negative, neutral) were assigned based on ratings, and texts were converted into TF–IDF features. Multinomial Naive Bayes was evaluated using 70:30 and 80:20 train–test splits. The best result was obtained at the 80:20 split achieving 71% accuracy. Class-wise metrics show very high positive recall (93.94%) but low negative recall (36.36%), indicating bias due to class imbalance. Naive Bayes outperforms KNN on this dataset due to its probabilistic model construction and faster prediction. Future work should use larger balanced data, apply sampling techniques, content-based labeling, and include multi-platform sources.