Optimasi Algoritma Support Vector Machine Berbasis Kernel Radial Basis Function (RBF) Menggunakan Metode Particle Swarm Optimization Untuk Analisis Sentimen

Abstrak

Di era digital, aplikasi Financial Technology (Fintech) telah menjadi bagian penting dalam kehidupan sehari-hari masyarakat. Kemudahan dan efisiensi yang ditawarkan oleh aplikasi Fintech menarik jutaan pengguna, yang aktif memberikan umpan balik dan ulasan di platform seperti Google Play Store. Ulasan ini menjadi sumber informasi berharga bagi pengembang untuk memahami persepsi pengguna, mengidentifikasi masalah, dan meningkatkan kualitas layanan. Penelitian ini bertujuan mengevaluasi efektivitas algoritma Particle Swarm Optimization (PSO) dalam meningkatkan akurasi analisis sentimen pada algoritma Support Vector Machine (SVM) dengan kernel Radial Basis Function (RBF). Data ulasan dikumpulkan melalui web scraping dari komentar di Google Play untuk tiga aplikasi Fintech, yaitu Flip, Neobank, dan Bank Jago. Tahapan pemrosesan meliputi pelabelan, preprocessing untuk membersihkan data, dan pembobotan kata menggunakan metode TF-IDF (Term Frequency-Inverse Document Frequency). Teknik Random Oversampling diterapkan untuk mengatasi ketidakseimbangan kelas dalam dataset. Hasil penelitian menunjukkan bahwa optimasi parameter dengan PSO mampu meningkatkan kinerja analisis sentimen, dengan peningkatan rata-rata sebesar 11,33% untuk setiap aplikasi. PSO juga meningkatkan akurasi model dalam menghadapi tantangan data tidak seimbang, memberikan wawasan yang lebih dalam bagi pengembang aplikasi untuk meningkatkan layanan.

Abstract

Financial technology (Fintech) applications have become part of people's daily lives in the digital era. The convenience and efficiency offered by Fintech applications have attracted millions of users, who actively provide feedback and reviews on platforms such as the Google Play Store. These reviews are an important source of information for application developers to understand user perceptions, identify problems, and improve service quality. The study investigates the effectiveness of the Particle Swarm Optimization (PSO) method for balanced and unbalanced datasets and how well it improves sentiment analysis accuracy when applied to the Support Vector Machine (SVM) algorithm when using Radial Basis Function (RBF) kernel. We conducted web scraping to collect user review data from Google Play for three FinTech applications: Flip, Neobank, and Bank Jago as research objects. Following data collection, the review data underwent preprocessing steps, such as word weighting using the TF-IDF (Term Frequency-Inverse Document Frequency), labeling, and preprocessing to clean the data. Random Oversampling resolved the dataset's class imbalance, making all classes representative in the study. The results of this study indicate that parameter optimization with PSO can improve the performance of sentiment analysis on the subjects studied. Furthermore, based on the results of SVM testing using parameter optimization of the PSO algorithm, an average performance increase of 11.33% was obtained for each application that had been analyzed. The results also show that PSO improves model accuracy in sentiment analysis with imbalanced data challenges, providing deeper insights for application developers to improve services.