Eko Yulianto, Nanang Nuryadi, Taufik Asra

Analysis of user satisfaction with the seabank app using support vector machines (SVM)

Introduction

Analysis of user satisfaction with the seabank app using support vector machines (svm). Analisis kepuasan pengguna aplikasi SeaBank dari ulasan Google Play Store menggunakan metode SVM. Studi ini mengungkap kenyamanan transaksi gratis, keluhan OTP, & masalah login.

Abstract

The development of financial technology (fintech) in Indonesia has driven the rapid growth of digital banks, one of which is SeaBank. User satisfaction with the SeaBank app is a crucial indicator for maintaining customer loyalty amid intense competition. This study aims to analyze user satisfaction with the SeaBank app based on reviews on the Google Play Store using the Support Vector Machine (SVM) method. Review data was classified into positive sentiment (satisfied) and negative sentiment (dissatisfied). The research stages included data collection, rating-based labeling, text preprocessing (including normalization of banking colloquial language), feature extraction using TF-IDF, handling imbalanced data with SMOTE, and classification using an RBF Kernel SVM. The results show that the application of SMOTE successfully addressed class imbalance, and SVM hyperparameter optimization yielded an accuracy of 89.14% with an F1-Score of 88.75%. The analysis findings indicate that the majority of users are satisfied with the convenience of free transactions, but there are significant complaints regarding OTP delays and errors in the login system.

Review

This study presents a timely and highly relevant analysis of user satisfaction with the SeaBank app, a critical endeavor in Indonesia's competitive digital banking sector. The authors effectively utilize Support Vector Machine (SVM) classification on Google Play Store reviews to categorize user sentiment, providing valuable insights into customer loyalty drivers. The methodology is commendably structured, encompassing essential steps such as robust text preprocessing, TF-IDF feature extraction, and the crucial application of SMOTE to address class imbalance—a common yet challenging aspect of real-world sentiment datasets. The reported high accuracy of 89.14% and an F1-Score of 88.75% underscore the robust performance of the optimized SVM model in discerning user satisfaction. While the research effectively identifies key satisfaction points, such as appreciation for free transactions, it also critically uncovers significant pain points like OTP delays and login system errors, offering actionable intelligence for SeaBank. To further enhance the study's depth, future iterations could provide more granular detail on the "rating-based labeling" process, specifying the exact threshold or criteria used to classify reviews as positive or negative based on star ratings. Additionally, while the RBF Kernel SVM's strong performance is evident, a brief discussion or comparison with alternative sentiment analysis models (e.g., deep learning or other machine learning classifiers) could provide broader context and reinforce the choice of SVM. Elaborating on the specific techniques employed for "normalization of banking colloquial language" would also highlight an interesting and potentially impactful aspect of the preprocessing. In conclusion, this paper delivers a valuable and pragmatic analysis for the fintech industry, demonstrating the efficacy of sentiment analysis in deriving actionable insights from user-generated content. The findings offer direct, data-driven recommendations for SeaBank to strategically enhance its app's user experience by prioritizing fixes for critical technical issues. Beyond SeaBank, the methodological framework presented is transferable and can serve as a robust model for other digital banks and fintech companies aiming to monitor and improve customer satisfaction. Future research could build upon this by incorporating a temporal analysis of reviews to track sentiment evolution or by conducting comparative studies across various digital banking platforms to identify overarching industry trends and benchmarks in user experience.

Full Text

Comments

You need to be logged in to post a comment.