Sentiment Analysis and Identification of Trust and Perceived Risk Aspects in User Reviews of the Dana Application Using Text Mining Methods and Support Vector Machine

Abstract

This research is motivated by the importance of understanding the trust and risk felt by users in using the DANA application based on user reviews. This research aims to analyze sentiment and identify the aspects of Trust and Perceived Risk in DANA application reviews using the Text Mining method and the Support Vector Machine (SVM) algorithm. The research data consists of 300 user reviews obtained from Kaggle. The research stages include pre-processing, sentiment labeling, label validation, word weighting using TF-IDF, and SVM classification. After eliminating neutral data, 280 data consisting of positive and negative sentiments were obtained, with 224 training data and 56 test data. The test results showed that SVM obtained accuracy, precision, recall, and F1-Score of 98% each. Identification of aspects using keyword dictionaries that have been validated by experts shows that Assurance is the most commonly found aspect of Trust, while Transaction Risk is the most dominant aspect of Perceived Risk. The results of the study recommend that developers maintain the Assurance aspect and improve the security and reliability of transactions to reduce the risks felt by users.