<![CDATA[The growth of digital lending services in Indonesia has contributed to a substantial increase in user reviews and complaints distributed across various online platforms, with Google Play Store being one of the most prominent. This condition poses a considerable challenge in the automatic detection of sentiment polarity, in line with the continuously growing volume of text data generated. This study aims to analyze user sentiment toward the Kredivo online lending application. The research methodology follows the SEMMA framework, which consists of five stages: Sample, Explore, Modify, Model, and Assess. Two classification algorithms were employed, namely Naïve Bayes and SVM, under two data splitting configurations of 80:20 and 70:30 for training and testing, respectively. Experimental results indicate that under the 80:20 configuration, Naïve Bayes achieved an accuracy of 92.01%, while SVM reached 97.05%. Under the 70:30 configuration, Naïve Bayes recorded an accuracy of 91.48% and SVM reached 96.76%. Evaluation using accuracy, precision, recall, and F1-score metrics confirmed that SVM consistently produced better classification performance compared to Naïve Bayes in categorizing user sentiment of the Kredivo online lending application. Based on the research results, it can be concluded that positive sentiment is more dominant than negative sentiment, with 6,120 reviews classified as positive and 2,012 reviews as negative.]]>
Copyrights © 2026
