Aspect-Based Sentiment Analysis On FLIP Application Reviews (Play Store) Using Support Vector Machine (SVM) Algorithm

Abstract

The development of fintech has driven the rapid growth of e-wallets like Flip, offering a convenient solution for interbank transfers without administrative fees. User reviews on the Play Store serve as crucial feedback for understanding the user experience. This research utilizes aspect-based sentiment analysis (ABSA) in combination with the SVM method to detect opinions, perceptions, and reviews pertaining to Flip's speed, security, and cost aspects. The objective is to provide valuable insights to both users and companies regarding their experiences with Flip in conducting financial transactions. The study employs a dataset comprising 13,500 preprocessed and cleansed data points, followed by TF-IDF vectorization. The data is divided into training and testing sets, utilizing techniques such as the train-test split and K-Fold Cross Validation to assess model performance. GridSearch analysis reveals that specific parameter combinations, notably C=1.0 and test_size=0.1, yield high accuracy across all aspects, with the linear kernel displaying the highest overall accuracy. Model evaluation is conducted using the confusion matrix and classification report, presenting accuracy, precision, recall, and F1-scores for each aspect. Notably, the Support Vector Machine model performs well, particularly in the speed, security, and cost aspects, where the cost aspect demonstrates exceptionally strong results. In summary, this study employs ABSA to analyze Flip application reviews, with the Support Vector Machine model showcasing impressive performance across various aspects, providing valuable insights for users and companies engaging with Flip's financial transaction services.Keywords: aspect-based sentiment analysis, support vector machine, reviews, Flip