Klasifikasi Sentimen Ulasan Aplikasi Spotify Di Google Play Store Menggunakan Algoritma C4.5

Abstract

Penelitian ini bertujuan untuk mengklasifikasikan sentimen pengguna aplikasi Spotify berdasarkan teks ulasan di Google Play Store dengan menggunakan algoritma C4.5. Sebanyak 298 ulasan dibagi secara seimbang antara sentimen positif dan negatif, yang sebelumnya telah diberi label berdasarkan rating ulasan. Proses pra-pemrosesan dilakukan melalui tahapan cleaning, case folding, tokenizing, stemming, dan stopword removal dengan bantuan aplikasi RapidMiner. Dengan menggunakan algoritma decision tree C4.5, model klasifikasi dibangun dan diuji berdasarkan metode cross validation. Hasil akhir penelitian menunjukkan bahwa model mampu mengklasifikasikan data dengan akurasi 81,56%, precision 90,92%, recall 71,03%, F1-score 79,03%, dan AUC sebesar 0,732. Temuan ini mengindikasikan bahwa C4.5 efektif dalam membedakan antara sentimen positif dan negatif pada teks ulasan berbahasa Indonesia. Penelitian ini memberikan kontribusi berupa evaluasi kinerja algoritma C4.5 pada analisis sentimen ulasan aplikasi digital, sehingga dapat menjadi dasar dalam memantau tingkat kepuasan pengguna serta mendukung pengambilan keputusan berbasis data ulasan sebuah aplikasi dan dapat menjadi referensi dalam pemilihan metode klasifikasi untuk data teks.

This research aims to classify user sentiment toward the Spotify application based on review text from the Google Play Store using the C4.5 algorithm. A total of 298 reviews were evenly divided between positive and negative sentiment, which had been previously labeled based on the review ratings. The preprocessing stage is carried out thru cleaning, case folding, tokenizing, stemming, and stopword removal steps with the help of the RapidMiner application. Using the C4.5 decision tree algorithm, a classification model is built and tested based on the cross-validation method. The final results of the study show that the model is able to classify data with an accuracy of 81.56%, precision of 90.92%, recall of 71.03%, F1-score of 79.03%, and an AUC of 0.732. This finding indicates that C4.5 is quite effective in distinguishing between positive and negative sentiment in Indonesian-language review texts. This study provides an empirical evaluation of the C4.5 algorithm for sentiment analysis of digital application reviews. The results may serve as a foundation for monitoring user satisfaction and supporting data-driven decision-making based on user reviews, as well as a reference for selecting suitable classification methods for textual datasets.