Sentiment Classification of AI Photo Studio: Pro Editor User Reviews on Google Play Store Using Multinomial Naïve Bayes

Abstract

This study analyzes the sentiment of Indonesian-language user reviews of the AI Photo Studio: Pro Editor application collected from the Google Play Store using Multinomial Naïve Bayes. A total of 2,028 unique reviews were obtained after duplicate removal, and 1,976 reviews remained after text preprocessing. The reviews were categorized into three sentiment classesnegative, positive, and neutralusing a lexicon-based labeling approach. The preprocessed reviews were represented using Term Frequency–Inverse Document Frequency (TF-IDF), divided into training and testing sets using an 80:20 stratified split, and balanced using the Synthetic Minority Over-sampling Technique (SMOTE) applied only to the training data. The Multinomial Naïve Bayes classifier was then trained and evaluated using accuracy, precision, recall, F1-score, and a confusion matrix. The final model achieved an accuracy of 75.00%. The negative class achieved a precision of 0.80, recall of 0.88, and F1-score of 0.84; the neutral class achieved a precision of 0.65, recall of 0.35, and F1-score of 0.46; while the positive class achieved a precision of 0.66, recall of 0.74, and F1-score of 0.70. The results indicate that the model performed best in classifying negative reviews, whereas the neutral class remained the most challenging category. These findings provide application-specific evidence of textual sentiment in AI-assisted photo-editing reviews and highlight the importance of interpreting written feedback alongside numerical ratings.