Analisis Sentimen Pada Ulasan Aplikasi Home Credit Dengan Metode SVM dan K-NN

Abstract

Dalam era teknologi, mencari pembiayaan finansial semakin mudah melalui aplikasi mobile seperti Home Credit. Aplikasi ini telah diunduh oleh lebih dari 10 juta pengguna Android dengan peringkat keseluruhan 4,4 di Google Play Store. Untuk membantu meninjau aplikasi, pengguna dapat memberikan ulasan dan penilaian di Google Play Store. Namun, dengan banyaknya ulasan, diperlukan analisis sentimen untuk mempermudah pemaha man. Dalam penelitian ini, dilakukan analisis sentimen menggunakan metode Support Vector Machine (SVM) dan K-Nearest Neighbor (KNN) pada data ulasan dari Google Play Store. Data yang diambil berjumlah 2.845 dengan informasi tentang skor dan komentar. Sentimen positif dan negatif ditentukan berdasarkan skor, dengan skor 4 dan 5 untuk sentimen positif, serta skor 1, 2, dan 3 untuk sentimen negatif. Setelah tahap preprocessing dan penghitungan tf-idf, dilakukan perhitungan menggunakan algoritma SVM dan KNN. Hasilnya menunjukkan bahwa metode SVM memiliki presisi 89%, recall 86%, F1-score 87%, dan akurasi 88%. Sementara metode KNN memiliki presisi 79%, recall 80%, F1-score 79%, dan akurasi 79%. Berdasarkan hasil tersebut, dapat disimpulkan bahwa metode Support Vector Machine lebih baik dalam melakukan analisis sentimen dalam penelitian ini.

In the age of technology, finding finance has never been easier through mobile apps like Home Credit. The app has been downloaded by over 10 million Android users with an overall rating of 4.4 on the Google Play Store. To help review the app, users can leave reviews and ratings on the Google Play Store. However, with so many reviews, sentiment analysis is needed to facilitate understanding. In this study, sentiment analysis using the Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) methods was conducted on review data from the Google Play Store. The data taken amounted to 2,845 with information about scores and comments. Positive and negative sentiments are determined based on scores, with scores of 4 and 5 for positive sentiments, and scores of 1, 2, and 3 for negative sentiments. After the preprocessing stage and tf-idf calculation, calculations are performed using the SVM and KNN algorithms. The results show that the SVM method has 89% precision, 86% recall, 87% F1-score, and 88% accuracy. While the KNN method has 79% precision, 80% recall, 79% F1-score, and 79% accuracy. Based on these results, it can be concluded that the Support Vector Machine method is better at performing sentiment analysis in this study.