Comparative Analysis of SARIMA and Holt-Winters Exponential Smoothing Methods for Rainfall Forecasting in Bandung City

Rainfall is an important factor that influences various social, economic, and environmental activities in Bandung City, so accurate forecasting is needed as a basis for disaster risk planning and mitigation. This study aims to analyze and compare the performance of the Seasonal Autoregressive Integrated Moving Average (SARIMA) and Holt-Winters Exponential Smoothing methods in forecasting monthly rainfall. The data used are monthly rainfall data for the period January 2016 to December 2024 sourced from the Bandung City Statistics Agency. The SARIMA analysis stages include stationarity checks, parameter identification and estimation, and diagnostic tests, while the Holt-Winters method is carried out through determining the type of model, component initialization, and smoothing parameter estimation. Evaluation of model accuracy uses the Mean Absolute Percentage Error (MAPE). The results show that the additive Holt-Winters method provides better performance with a MAPE value of 64,410%, lower than the SARIMA model which produces a MAPE of 76,787%, so it is selected as the best method. The 2025 forecast shows a rainfall pattern that tends to be high at the beginning and end of the year and decreases in the middle of the year, in line with relatively consistent historical seasonal patterns.