dc.description.abstract | Floods are natural events that have erratic pattern. The time when floods occure are detectable, when we are able to forecast rainfall. The amount of rainfall is influenced by several factors, including temperature and humidity. Rainfall, temperature, and humidity are time series data that have are not independent. Forecasting for time series data can be done by using ARIMA model. Forecasting result from ARIMA models are not satisfying because it is not close to the actual data, which was also showed by its MAPE value of 37.01%. Hence the transfer function model was applied to forecast rainfall as the output variable along with temperature and humuidity factor as the input variable. The result from transfer function model used for short run forecasting (1 week) was quite good which can be seen by its MAPE value of 5.28%. Although transfer function model has a lower MAPE value compared to ARIMA model, transfer function model is not good enough to be used on long-term (1 month) rainfall forecasting, which is showed by its MAPE value of 31.68% and it could not detect flood occurence well. Based on those MAPE values it can be concluded that the transfer function model is still better than ARIMA model. | en |