Comparison of Rating-based and Inset Lexicon-based Labeling in Sentiment Analysis using SVM (Case Study: GoBiz Application Reviews on Google Play Store)

Abstract

Digital transformation has impacted various sectors, including Micro, Small, and Medium Enterprises (MSMEs). GoBiz, a partner platform for Gojek's GoFood service, plays a crucial role in supporting MSME digitalization, making it essential to understand user perceptions of the application. This study conducts sentiment analysis on 5,000 GoBiz user reviews from the Google Play Store. It compares two labeling methods—Rating-Based and Inset Lexicon—and evaluates them using the Support Vector Machine (SVM) algorithm. The analysis process includes data selection, text preprocessing, data transformation using TF-IDF, SVM implementation with 10-fold cross-validation, and result visualization through word clouds. The findings indicate that the Rating-Based labeling method achieved an accuracy of 87%, with a precision of 86.7%, recall of 87.1%, and an F1-score of 86.8%. Meanwhile, the Inset Lexicon labeling method outperformed it, achieving an accuracy of 89.7%, precision of 89%, recall of 89.8%, and an F1-score of 89.3%. These results suggest that the combination of the Inset Lexicon labeling method and the SVM algorithm is more effective in classifying user sentiment and providing a more accurate understanding of user perceptions regarding the GoBiz application. Sentiment analysis results indicate that users appreciate GoBiz’s ease of operation but face challenges with driver services and advertisement features, highlighting areas for improvement to enhance user satisfaction.