Analysis of Classification Algorithm Performance on User Review Sentiment of the Muamalat DIN Application

Abstract

Banking applications have become an integral part of modern society. One such application is Muamalat DIN, launched by Bank Muamalat Indonesia with the aim of facilitating customers in conducting various transactions and activities. User reviews of this application vary widely, ranging from positive to negative comments. The purpose of this study is to evaluate user attitude on reviews of Bank Muamalat Indonesia's digital banking product, the Muamalat DIN application. This research offers insights into the efficacy of the SMOTE balancing technique compared to undersampling by utilizing a methodology that includes data collection via scrapping techniques, data preprocessing, and the application of Multi Layer Perceptron (MLP), XGBoost, and LightGBM classification algorithms. The results show that SMOTE-paired XGBoost works better for sentiment categorization. The study's conclusion emphasizes the significance of choosing the right data balancing method to increase sentiment analysis's accuracy in Islamic banking applications, which can be used as a foundation for strategies aimed at enhancing customer service and making decisions.