Sentiment Analysis of User Reviews for the Locket Widget Application on Google Play Store Using the Naive Bayes Classifier

Abstract

User reviews on the Google Play Store contain valuable opinions that can be utilized to evaluate application quality, including the Locket Widget application. This study aims to classify user review sentiment into positive, neutral, and negative categories using the Naïve Bayes Classifier algorithm with Term Frequency–Inverse Document Frequency (TF-IDF) weighting. The data were collected through web scraping, resulting in 589 user reviews. The dataset then underwent preprocessing, sentiment labeling, TF-IDF weighting, and sentiment classification. The model was evaluated using an 80:20 stratified train–test split. The results showed that the proposed model achieved an accuracy of 76.27%, with a weighted precision of 0.69, weighted recall of 0.76, and weighted F1-score of 0.72. The findings indicate that the combination of TF-IDF and the Naïve Bayes classifier is effective in classifying positive and negative user review sentiments. However, the model was unable to effectively classify the neutral class, resulting in an F1-score of 0.00 for this category. This finding indicates that further improvements are needed to enhance the model's ability to distinguish neutral sentiment from positive and negative classes.