SARIMA and Holt-Winters Seasonal Methods for Time Series Forecasting in Tuberculosis Case

People often deal with health problems and illnesses in their lives. One of the deadly infectious diseases is Tuberculosis. The high number of cases compared to the reduction target is quite a serious problem. A method is needed to predict future data uncertainty to alert to the danger of advanced Tuberculosis cases. The research steps are 1) data preparation, 2) data training process, 3) data testing or validating and evaluation process; and 4) Forecasting. Modeling is done using the Seasonal ARIMA(SARIMA), Additive Holt Winter (AHW), and Multiplicative Holt Winter (MHW) methods. Data Scenario is divided into 70% for training data, 30% for testing data, or validation of the resulting model accuracy. The performance evaluation model is checked by comparing the Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE) error values between the actual and forecast data. According to the data processing scenario, the Multiplicative Holt Winter (MHW) method has the RMSE value of 124.60 and MAPE of 0.13 or the lowest compared to others.