Perbandingan Kinerja Model Prediksi Cuaca: Random Forest, Support Vector Regression, dan XGBoost

Abstract

Accurate weather predictions are essential to mitigate the impacts of weather changes and support better planning in sectors such as agriculture, transportation, and tourism. Indonesia often faces unpredictable weather, such as sudden rains and long droughts, which can cause huge losses. This study aims to compare the performance of three machine learning algorithms Random Forest, Support Vector Regression (SVR), and XGBoost in predicting weather using meteorological data (minimum temperature, maximum temperature, rainfall, wind direction, average humidity) as well as IoT data totaling 1650 data per variable. The variables used in this study include minimum temperature, maximum temperature, rainfall, wind direction, and average humidity. Data analysis techniques were performed using three main evaluation metrics, namely Mean Absolute Error (MAE), Mean Squared Error (MSE), and R-squared (R²). The results showed that XGBoost gave the best performance with MAE 0.3744, MSE 0.2278, and R² 0.8183. Random Forest and SVR also produced good predictions, with MAE values of 0.3869 and 0.3820, MSE 0.2422 and 0.2524, and R² 0.8068 and 0.7987, respectively. The results show XGBoost is the best model for weather prediction, which can help improve accuracy in agricultural planning and weather-related disaster risk mitigation.