ANALISIS SENTIMEN ULASAN PENGGUNA TERHADAP APLIKASI NO THANKS! DI GOOGLE PLAY STORE MENGGUNAKAN METODE NAIVE BAYES

Lingga; Firman Noor Hasan;

Analisis Sentimen Ulasan Pengguna Terhadap Aplikasi No Thanks! Di
Google Play Store Menggunakan Metode Naive Bayes

Lingga

Penelitian ini bertujuan untuk menganalisis sentimen pengguna terhadap aplikasi No Thanks! di
Google Play Store menggunakan metode Naive Bayes. Aplikasi ini menjadi sorotan karena
fungsinya sebagai panduan etis bagi pengguna dalam mendukung gerakan boikot. Sebanyak 2.000
data ulasan dikumpulkan melalui teknik web scraping dan diproses menggunakan tahapan text
preprocessing yang mencakup cleansing, case folding, tokenizing, stopword removal, dan stemming.
Selanjutnya, data diubah menjadi bentuk numerik menggunakan metode TF-IDF dan diberi label
sentimen berdasarkan skor rating ulasan. Model klasifikasi dilatih menggunakan algoritma
Multinomial Naive Bayes untuk mengkategorikan ulasan ke dalam tiga kelas sentimen: positif,
netral, dan negatif. Evaluasi dilakukan menggunakan metrik accuracy, precision, recall, dan F1-
score. Hasil penelitian menunjukkan bahwa sebagian besar ulasan bersentimen positif, dengan
akurasi model mencapai 92%. Namun, model mengalami kesulitan dalam mengenali sentimen netral
dan negatif akibat distribusi data yang tidak seimbang. Hasil ini menunjukkan persepsi dominan
pengguna terhadap aplikasi serta memberikan wawasan mengenai efektivitas model dalam konteks
sentiment analysis berbasis opini publik.
Kata Kunci: analisis sentimen, Naive Bayes, No Thanks!, Google Play Store, TF-IDF
Sentiment Analysis of User Reviews on the No Thanks! Application in Google

Play Store Using the Naive Bayes Method

Lingga

This study aims to analyze user sentiment toward the No Thanks! application on Google Play Store
using the Naive Bayes method. The app is notable for its role as an ethical guide aligned with boycott
movements. A total of 2,000 user reviews were collected via web scraping and processed through
text preprocessing stages, including cleansing, case folding, tokenizing, stopword removal, and
stemming. The data were then transformed into numerical form using TF-IDF and labeled based on
rating scores. The classification model was trained using the Multinomial Naive Bayes algorithm to
categorize reviews into positive, neutral, and negative sentiments. Evaluation was conducted using
accuracy, precision, recall, and F1-score metrics. Results showed that most reviews were positive,
with the model achieving 92% accuracy. However, performance was limited on neutral and negative
classes due to data imbalance. These findings reflect users' dominant perceptions of the app and
provide insight into the effectiveness of sentiment analysis models applied to public opinion data.
Keywords: Naive Bayes, Multinomial, user reviews, No Thanks!, Google Play Store.