Electronic Theses and Dissertation

Analisis sentimen digunakan untuk dapat memberikan wawasan kepada pemberi
layanan agar dapat meningkatkan kualitas layanan yang diberikan. Ulasan-ulasan
pengguna terkait kepuasan mereka terhadap suatu layanan diperoleh pada
platform Google Play Store.
Penelitian ini melakukan analisis sentimen
menggunakan tiga algoritma machine learning, yaitu Random Forest, Decision
Tree, Logistic Regression terhadap 5 aplikasi BUMN pada tahun 2025. Ulasan
pengguna akan dikelompokkan ke dalam sentimen positif, negatif dan netral.
Algoritma Random Forest menunjukkan nilai akurasi yang tinggi pada data
pelatihan, meskipun pada beberapa objek cenderung overfitting. Dan Decision
Tree menunjukkan overfitting yang lebih parah dibanding Random Forest.
Sementara itu, Logistic Regression menghasilkan performa paling stabil yang
memiliki selisih yang kecil antara data pelatihan dan testing, menjadikan
algoritma yang lebih baik untuk generalisasi prediksi sentimen pada data baru.
Hasil analisis, Logistic Regression direkomendasikan untuk analisis sentimen
pada dataset ini, sedangkan Random Forest dapat digunakan jika memerlukan
nilai akurasi yang tinggi untuk data pelatihan.

Sentiment analysis is utilized to provide insights for service providers to improve the quality of services offered. User reviews regarding their satisfaction with a given service were collected from the Google Play Store platform. These user reviews were classified into three sentiment categories: positive, negative, and neutral. This study conducts sentiment analysis using three machine learning algorithms—Random Forest, Decision Tree, and Logistic Regression—on reviews from five state-owned enter prise (BUMN) applications in 2025. The results indicate that the Random Forest algorithm achieves high accuracy on the pelatihan data, although it tends to overfit in several cases. The Decision Tree algorithm exhibits even more pronounced overfitting compared to RandomForest. In contrast, Logistic Regression demonstrates the most stable performance, with a minimal gap between pelatihan and pengujianaccuracy, making it more suitable for generalizing sentiment predictions on unseen data. Based on the analysis, Logistic Regression is recommended for sentiment classification on this dataset, while Random Forest may be preferred in scenarios that prioritize high pelatihan accuracy.