Modeling the Reputation of Digital Banks Based on Public Opinion Using a Text Mining Approach with TF-IDF and the Support Vector Machine (SVM) Algorithm

Abstract

This study addressed the growing importance of reputation management in digital banking, where public opinion expressed on social media significantly influences customer trust and business sustainability. The objective of this research was to model the reputation of a digital bank based on public sentiment using a text mining approach. The study employed the CRISP-DM methodology, including data collection, preprocessing, modeling, and evaluation. A total of 1,897 Twitter comments related to the "Jenius" digital banking application were collected from 2023 to 2025. The data underwent preprocessing stages such as case folding, cleansing, tokenizing, normalization, stopword removal, negation handling, and stemming. Feature extraction was performed using Term Frequency–Inverse Document Frequency (TF-IDF), and sentiment classification was conducted using Support Vector Machine (SVM). The performance of SVM was compared with Naïve Bayes and K-Nearest Neighbors (KNN). The results showed that SVM achieved the best performance with an accuracy of 81.58%, outperforming Naïve Bayes (70.26%) and KNN (55.00%). Furthermore, sentiment distribution indicated that positive sentiment dominated public opinion, reflecting a generally favorable perception of the digital bank. In conclusion, the combination of TF-IDF and SVM proved effective for sentiment classification and can be utilized to model digital bank reputation, providing valuable insights for improving service quality and customer satisfaction.