Dengue hemorrhagic fever (DHF) is endemic in Banyumas Regency, where case patterns may reflect both temporal trends and climate variability. This study developed a hybrid forecasting model combining Autoregressive Integrated Moving Average (ARIMA) and Support Vector Regression (SVR), with rainfall, ambient temperature, and relative humidity as hydrometeorological predictors. The dataset comprised 36 monthly observations from January 2022 to December 2024, obtained from health surveillance records of the Banyumas Regency Health Office and climate data from BMKG and BPS Banyumas. Climate variables were incorporated with one- and two-month lags to account for potential delayed associations with case counts. Model performance was evaluated using out-of-sample data from May to December 2024. The hybrid ARIMA–SVR model achieved an RMSE of 17.11, an MAE of 11.67, and a MAPE of 50.94%. Its RMSE and MAE were lower than those of standalone ARIMA(1,0,1) (17.85 and 12.40, respectively) and standalone SVR (25.72 and 17.81, respectively). Although the hybrid model performed better than the individual models on these error measures, its MAPE indicates substantial relative prediction error. The findings suggest that combining time-series and climate information may support dengue forecasting; further evaluation with longer time series and additional validation periods is needed before operational use as an early-warning tool.
Keywords
dengue hemorrhagic fever, Banyumas Regency, ARIMA, support vector regression, hybrid forecasting, hydrometeorological predictors