Prediksi Kepuasan dan Loyalitas Mahasiswa Menggunakan Random Forest dengan Pendekatan Explainable AI Berbasis SHAP

Indah Kusuma Dewi (Program Studi Teknik Informatika, Fakultas Sains dan Teknologi, Universitas Ibnu Sina, Batam)
Muhammad Ropianto (Program Studi Teknik Informatika, Fakultas Sains dan Teknologi, Universitas Ibnu Sina, Batam)
Elsha Dellista Ashofie (Program Studi Teknik Informatika, Fakultas Sains dan Teknologi, Universitas Ibnu Sina, Batam)

JiTEKH (Jurnal Ilmiah Teknologi Harapan)
Vol. 14 No. 2 (2026): September 2026

Student satisfaction and loyalty are two strategic indicators of higher education institutions' success in maintaining educational service quality. This study builds prediction models for student satisfaction and loyalty using the Random Forest algorithm, interpreted transparently through the SHapley Additive exPlanations (SHAP) approach. Data were obtained from 437 student respondents covering eight service dimensions with 25 indicator variables. The satisfaction model achieved 94.32% accuracy (AUC = 0.8819), while the loyalty model obtained 87.50% accuracy (AUC = 0.5490). SHAP analysis identified Student financial difficulty solutions, Campus Wi-Fi stability, and Variation in teaching methods as dominant predictors of satisfaction, while Scholarship transparency is the main driver of loyalty.