COMBINATION OF LOGISTIC REGRESSION AND NAÏVE BAYES IN SENTIMENT ANALYSIS OF ONLINE LENDING APPLICATION PLATFORMS BY UTILIZING THE LEXICONS FEATURE

In the digital age, online lending apps have become an important tool in facilitating financial transactions and supporting MSMEs. However, the existence of negative opinions related to violations such as theft of customer data raises concerns in the community. This research aims to analyze sentiment towards online loan applications, especially Kredivo, using a combination of Logistic Regression and Naïve Bayes which is optimized through the Lexicons feature. Data is taken from Google Play Store reviews, then labeling, preprocessing, and feature extraction are executed through TF-IDF technique. The classification models built are Naive Bayes (NB) and Logistic Regression (LR), where the results of the two models are combined with the ensemble voting method using lexicons features. The evaluation results show that the combination approach of the three methods can significantly improve classification accuracy compared to the use of a single method. The combined model achieved an accuracy of 89.62%, higher than Logistic Regression (86.19%) and Naive Bayes (83.54%).