<![CDATA[Purpose: This study examines recent Shopee mobile application reviews from a service diagnostic perspective by linking sentiment classification with aspect-level interpretation. The aim is to identify not only whether user feedback is positive or negative, but also which service dimensions are associated with satisfaction and dissatisfaction. Methods: A total of 160,492 Google Play Store reviews of the Shopee Android application were collected from the 2026 review period. After cleaning, rating-based sentiment labeling, and preprocessing, a stratified random sample of 50,000 reviews was used for modeling. TF-IDF unigram and bigram features were applied with a majority class baseline, Naive Bayes, Logistic Regression, and Linear SVM. Performance was evaluated using accuracy, Macro F1-score, confusion matrix, classification report, and 5-fold stratified cross-validation. A keyword-based aspect mapping was then used to interpret sentiment across six service dimensions. Result: The TF-IDF and Logistic Regression model achieved 0.9194 accuracy, 0.9096 Macro F1-score, and 0.899 cross-validated Macro F1-score. Aspect-level findings show that positive sentiment was concentrated in Promotion and Voucher, Product and Seller, and Customer Service, while negative sentiment was concentrated in Payment, Delivery, and Application Performance. Advertising and delivery-related expressions emerged as key negative signals. Novelty: This study proposes a service diagnostic task with aspect-level analysis, which includes sentiment classification, term-weight interpretation, and service-aspect mapping, to position Shopee review analysis. The strategy transforms bulk app reviews into actionable data to track platform service quality.]]>
Copyrights © 2026
