<![CDATA[Abstrak - Pengelolaan data kepuasan pasien menjadi komponens yang semakin penting dalam meningkatkan kualitas pelayanan kesehatan di era digital. Penelitian ini bertujuan untuk mengklasifikasikan tingkat kepuasan pasien di Puskesmas Mootilango dengan menerapkan algoritma Extreme Gradient Boosting (XGBoost) berbasis dimensi SERVQUAL. Tahapan penelitian dimulai dari pengumpulan data kuesioner, preprocessing, hingga implementasi model melalui optimasi Gradient dan Hessian. Transparansi logika algoritma divalidasi secara manual melalui perhitungan nilai Similarity Score dan Gain. Hasil penelitian menunjukkan bahwa model XGBoost memiliki performa andal dengan tingkat akurasi sebesar 80,00% dan recall sebesar 88%. Analisis Feature Importance mengidentifikasi bahwa indikator Q5 (Tata Letak Ruangan) dari dimensi Tangibles dan Q8 (Keakuratan Diagnosa) dari dimensi Reliability merupakan faktor paling berpengaruh terhadap kepuasan pasien. Simulasi manual pada indikator Q5 menghasilkan nilai Total Gain sebesar 0,7238, yang membuktikan kontribusi signifikan fitur tersebut dalam meningkatkan kemurnian klasifikasi. Disimpulkan bahwa algoritma XGBoost efektif memberikan rekomendasi strategis bagi manajemen untuk memprioritaskan perbaikan pada aspek fisik dan keandalan medis demi optimalisasi kualitas pelayanan. Kata kunci : Kepuasan Pasien; XGBoost; SERVQUAL; Machine Learning; Feature Importance; Abstract -Patient satisfaction data management is becoming an increasingly important component in improving the quality of healthcare services in the digital era. This study aims to classify patient satisfaction levels at the Mootilango Community Health Center by applying the Extreme Gradient Boosting (XGBoost) algorithm based on the SERVQUAL dimension. The research stages start from questionnaire data collection, preprocessing, to model implementation through Gradient and Hessian optimization. The transparency of the algorithm logic is manually validated by calculating the Similarity Score and Gain values. The results show that the XGBoost model has reliable performance with an accuracy rate of 80.00% and a recall of 88%. Feature Importance analysis identified that indicator Q5 (Room Layout) from the Tangibles dimension and Q8 (Diagnostic Accuracy) from the Reliability dimension are the most influential factors on patient satisfaction. Manual simulation on indicator Q5 produced a Total Gain value of 0.7238, which proves the significant contribution of this feature in improving the purity of the classification. It was concluded that the XGBoost algorithm effectively provides strategic recommendations for management to prioritize improvements in physical aspects and medical reliability to optimize service quality. Keywords: Patient Satisfaction; XGBoost; SERVQUAL; Machine Learning; Feature Importance;]]>
