<![CDATA[Traditional evaluation of Bahasa Indonesia for Foreign Speakers (BIPA) learning success relies on subjective teacher assessments lacking objectivity. This study aims to integrate DBSCAN and XGBoost algorithms to analyze learning patterns and dominant success factors in BIPA. Quantitative Educational Data Mining (EDM) exploratory approach applied to 200 students population at Songsermsasana School, Hat Yai, Thailand during 27-day KKN, using complete tabular data sample. Instruments include activity scores, class participation, attendance, exam/quiz scores, study time, and question frequency variables; analysis techniques involve preprocessing, DBSCAN clustering (Īµ=0.5, minPts=5), and XGBoost feature importance. Results reveal three clusters: Cluster 0 (high speaking/writing >80), Cluster 1 (stable receptive skills), Cluster 2 (low attendance), with speaking score (15.89%) and writing score (11.61%) dominant. Hybrid model outperforms K-Means in handling noise. Research provides objective data-driven evaluation for global BIPA teaching personalization.]]>
Copyrights © 2026
