PREDIKSI KASUS BUNUH DIRI MENGGUNAKAN ALGORITMA XGBOOST DAN TEKNIK HYPERPARAMETER TUNING

Suicide is a global health problem that causes more than 800,000 deaths each year. This figure shows that suicide is not only an individual problem, but can also be influenced by various factors such as social, economic, and mental health conditions. Therefore, a data-based approach is needed that can accurately predict the number of suicide cases. This study aims to build a model to predict the number of suicide cases using the XGBOOST algorithm, with
improved performance through hyperparameter tuning techniques using RandomSearchCV. The dataset used is sourced from the WHO for the period 1985–2016, covering demographic and socioeconomic features. The model is evaluated using MAE, MSE, and R-squared metrics. The results show that the XGBOOST model has excellent overall prediction performance, with high R-squared values across all continents. The model is able to gradually improve predictions through the use of residual values, where from the initial iteration the prediction results are already close to the actual values in each region. The highest R-squared value was achieved in Asia at 99.78%, while the lowest prediction error was obtained in Africa (MAE 1.74, MSE 16.13). These findings indicate that a machine learning-based approach and optimal parameter
tuning can contribute to supporting data-driven suicide prevention efforts.