STUDY OF HYBRID FLOOD FORECASTING APPROACH COMBINING MULTIPLICATE SEASONAL ARIMA AND HYBRID- NEURO FUZZY BASED ON LONG-TERM TIME SERIES

Author

Sri Supatmi, S.Kom.,M.T., D.Sc Irfan Dwiguna Sumitra, S.Kom., M.Kom. Ph.D

Abstrak

This research proposes a modern hybrid method to forecast the flood employing an approach combining Multiplicative Seasonal Autoregressive Moving Average (MSARIMA) and Hybrid-Neuro Fuzzy Inference System (HN-FIS) based on long-term time series. This proposed method is called Hybrid Flood Forecasting System Technology (Hybrid-FFST). This research aims to improve three previous types of research on flood forecasts, such as flood prediction using HNFIS and flood forecasting using MSARIMA and rainfall prediction using Multiplicative Seasonal Autoregressive Moving Average Subsequence Aggregate (MSARIMASA). This research has taken place in Bandung West Java Province, Indonesia. The performance of flood event forecasting improving by using the hybrid approaches method using both MSARIMA and two levels of HN-FIS. This proposed method employs six parameters: rainfall, temperature, population density, large watershed, the altitude of the area, and slop of the land to predict the flood event. The performance of this method is generated and fitting well using the Hybrid-FFST approach and the verified by Mean Absolute Percentage Error (MAPE), Root Means Square Percentage Error (RMSPE), and Mean Forecast Error (MFE) to identify the best-fitted model of the proposed model. The proposed model's performance is compared using MAPE, RMSPE, and MFE with MSARIMA, HN-FIS, and MSARIMASA model. The confidence performance of the proposed method obtains more significant than 97% according to the MAPE, RMSE, and MFE values. The proposed model results indicate better performance than the MSARIMA, HN-FIS, and MSARIMASA models to forecast the flood event. The impact of this research is the flood can be predicted before occurred in someplace.