Performance Comparison of Random Forest and XGBOOST Algorithms on E-Commerce App Sentiment Analysis

Abstract

The rapid advancement of digital technology has driven significant growth in e-commerce applications in Indonesia, with Shopee emerging as one of the leading platforms boasting millions of active users. User reviews on the Google Play Store serve as a crucial data source for sentiment analysis and evaluating the quality of app services. However, the sheer volume and linguistic complexity of these reviews demand efficient and accurate analytical methods. This study aims to compare the performance of the Random Forest and XGBoost algorithms in classifying the sentiment of Shopee user reviews. Review data were collected through web scraping from the Google Play Store and processed using several labelling, casefolding, cleaning, tokenizing, stopword removal, and stemming. Feature extraction was performed using the TF-IDF method, and the data were split into training and testing sets. Sentiment classification models were built using the Random Forest and XGBoost algorithms and evaluated based on accuracy, precision, recall, and F1-score metrics. The results of this study are expected to provide recommendations for the most effective algorithm for sentiment analysis in e-commerce applications and contribute to the development of services and further research in data mining and natural language processing.