Identifikasi Tingkat Stres Mahasiswa Menggunakan Algoritma XGBoost dengan Data Psikologis, Fisik, Akademik, dan Lingkungan

Abstract

Stress is a health problem widely experienced by students due to interrelated and multidimensional academic, psychological, physical, and environmental demands, making conventional identification approaches inadequate for revealing complex patterns. This study applies the Extreme Gradient Boosting (XGBoost) algorithm to identify student stress levels based on the StressLevelDataset, evaluate the performance of the algorithm, and identify the most influential factors in the classification process. The model was built in two stages, namely a baseline model with default parameters and a model resulting from hyperparameter tuning using Grid Search. The results show that the tuned XGBoost model, evaluated using Stratified 10-Fold Cross-Validation, achieved an accuracy of 88.91%, a precision of 88.95%, a recall of 88.92%, and an F1-score of 88.93%, an improvement over the baseline model (87.27% accuracy). Feature importance analysis revealed that blood_pressure, sleep_quality, and safety were the three most influential variables based on gain values, with blood_pressure as the most dominant factor (±45.6%). These results indicate that the XGBoost algorithm is effective and reliable for objectively identifying student stress levels based on data