Abstract: This paper applies a variable fuzzy set method based on entropy weight to construction of a
framework of classified real-time flood forecasting using the concepts of clustering and classification. Fuzzy
clustering was used to classify historical floods based on the flood antecedent impact factors and
time-varying rainfall information. In the present study, this conceptual hydrological model was calibrated for
each type of flood. In application, a fuzzy diagnosis model was used to identify the types of floods in future
by using the flood information obtained. Results show that the classified framework gives a fast and accurate
diagnosis of the type of flood and significantly improves the accuracy of flood forecasting.