Rainfall Prediction Using Random Forest with Synthetic Minority Over-Sampling Technique (SMOTE)
TL;DR
A rainfall prediction model based on machine learning using the Random Forest algorithm optimized with the Synthetic Minority Over-Sampling Technique (SMOTE) to address data imbalance has potential as a decision-support tool for weather-risk mitigation at Tanjung Perak Port.
Abstract
Rainfall is a crucial meteorological parameter that directly affects maritime activities and port operations, particularly in coastal regions. Tanjung Perak Port, as one of the busiest ports in Indonesia, is highly vulnerable to weather disturbances due to rainfall variability. This study develops a rainfall prediction model based on machine learning using the Random Forest algorithm optimized with the Synthetic Minority Over-Sampling Technique (SMOTE) to address data imbalance. Daily meteorological data for 2013-2024 were obtained from NASA POWER, including rainfall, minimum and maximum temperature, relative and specific humidity, wind speed, and surface pressure. Model performance was evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). SMOTE improved the model's ability to identify high-intensity rainfall events. The combined 2024-2025 evaluation produced an MAE of 1.457 mm, an RMSE of 1.819 mm, and R² of 0.861. The model therefore captures seasonal rainfall patterns and has potential as a decision-support tool for weather-risk mitigation at Tanjung Perak Port.