ANALISIS KEPUASAN PENGGUNA APLIKASI OCTO MOBILE MENGGUNAKAN SUPPORT VECTOR MACHINE (SVM)

Abstract

Digital transformation has driven innovation in banking services, exemplified by the Octo Mobile application launched by PT Bank CIMB Niaga Tbk. User reviews on the Google Play Store serve as a rich source of unstructured data that can be leveraged to measure user satisfaction in real-time. This study aims to analyze user satisfaction sentiment regarding the Octo Mobile application post-rebranding using an optimized Support Vector Machine (SVM) algorithm. The research methodology adopts a text mining approach, employing web scraping techniques to collect 8,000 user reviews from the period of January to December 2023. The dataset exhibited a significant class imbalance (85.56% positive and 14.44% negative). To address this issue, the Synthetic Minority Over-sampling Technique (SMOTE) was implemented. As a novel approach, this study applied a customized banking colloquial normalization dictionary during the pre-processing stage to handle slang and transactional abbreviations commonly found in reviews. Feature extraction was conducted using Term Frequency-Inverse Document Frequency (TF-IDF), followed by SVM classification with hyperparameter optimization via Grid Search. The test results demonstrated that the SVM model with an RBF kernel (C=10, γ=0.1) achieved an accuracy of 88.18% and an F1-Score of 87.93% for the negative class, successfully surpassing the 85% accuracy target and outperforming previous studies. The final analysis revealed that 85.56% of users expressed satisfaction with the user interface and transaction ease, whereas 14.44% expressed dissatisfaction, predominantly regarding technical issues such as lagging, system errors, and OTP delays. These findings prove that the combination of SMOTE, contextual normalization, and SVM optimization yields a valid and unbiased analysis, providing strategic recommendations for bank management.