SENTIMENT ANALYSIS APPROACH TO USER PERCEPTIONS OF AMMANA SHARIA FINTECH BASED ON MACHINE LEARNING ALGORITHMS

Abstrak

This study aims to analyze and classify user sentiments toward the Islamic fintech application Ammana based on user reviews collected from the Google Play Store. The data were obtained through a web scraping process and underwent text preprocessing, including case folding, stopword removal, stemming, and tokenization using the Sastrawi library in Python through Google Colab. The research employed a quantitative comparative approach by comparing three machine learning algorithms: Support Vector Classifier (SVC), Multinomial Naive Bayes (MNB), and Random Forest (RF). The classification results indicate that SVC achieved the best performance with an accuracy of 91.80%, precision of 92.67%, recall of 91.80%, and F1-score of 91.99%. Meanwhile, Random Forest ranked second with an accuracy of 87.98%, and Naive Bayes performed the lowest with an accuracy of 85.24%. Overall, the findings suggest that SVC is the most effective algorithm for identifying positive, negative, and neutral user sentiments toward the Ammana application. From a business and financial technology perspective, sentiment analysis serves as a strategic tool for monitoring public perception, enhancing user experience, and strengthening customer trust in digital Islamic financial services.