User Sentiment Analysis Towards Online Loan Services Based on Social Media Data X Using the Bert Method

Abstract

Shopee PayLater (SPayLater) is an online lending service widely used by Indonesians. The high usage of this service has led to diverse user opinions regarding the experience of using SPayLater on social media platform X. The large number of comments generated makes the manual analysis process inefficient, so an automated method capable of quickly and accurately recognizing user sentiment tendencies is needed. This study aims to analyze user sentiment towards the SPayLater service and evaluate the performance of the IndoBERT model in classifying comments based on social media platform X data. The research data consists of 800 comments collected manually using keywords related to SPayLater. Next, the data went through preprocessing stages including data cleaning, case folding, tokenizing, and normalization before being divided into training data and test data for the IndoBERT model fine-tuning process. The fine-tuned IndoBERT model achieved an accuracy of 97.50%, a precision of 97.56%, a recall of 97.50%, and an F1-score of 97.46%. In the classification results, negative sentiment dominated with 392 comments, followed by neutral sentiment with 293 comments and positive sentiment with 115 comments. This achievement proves that IndoBERT is capable of classifying user sentiment with a very high level of accuracy, making this method suitable for sentiment analysis of online lending services based on social media data.