FIDs Classifier for Artificial Intelligence and Its Application

摘要

Fuzzy ID3 (FIDs) is popular and efficient method of making fuzzy decision trees in recent years. This paper presents FIDs algorithm for the precipitations during typhoon periods for a reservoir watershed. The FIDs was constructed as the quantitative precipitation forecast (QPF) model. This study also constructed the traditional C4.5 and the average statistical model (AVS) to compare with the performance by FIDs model. The steps involve collecting typhoon data, preprocessing the typhoon patterns, building QPF models, and training and testing the models. The experiment was in Shihmen Reservoir watershed. The results include the analysis of the 1-, 3-, and 6-hr accumulated rainfalls. The results showed that the superior RMSE and the categorical statistics of BIAS and ETS scores by using FIDs in contrast to those by using traditional C4.5 and AVS. Consequently, the FIDs model demonstrated its feasibility for predict rainfalls.