In order to assess hydrologic and nonpoint source pollutant behaviors in the watershed with the Soil and Water Assessment Tool (SWAT) model, accuracy evaluation of the SWAT model should be conducted prior to the application of it to the watershed. When calibrating and validating hydrological components of the SWAT model, the Nash-Sutcliffe efficiency coefficient (EI) has been widely used. However, the EI value has been known as it is affected sensitively by big number among a range of numbers. In this study, the Web-based flow clustering EI estimation system using K-means clustering algorithm was developed and used for the SWAT hydrology evaluation. Even though the EI of total streamflow was high, the EI values of hydrologic components (i.e., direct runoff and baseflow) were not high. Also when the EI values of flow group I and II (i.e., low and high value group) clustered from direct runoff and baseflow were computed, respectively, the EI values of them were much lower, negative EI values for some flow group comparison. The SWAT auto-calibration tool estimated values also showed negative EI values for most flow group I and II of direct runoff and baseflow although EI value of total streamflow was high. The result obtained in this study indicates that the SWAT hydrology component should be calibrated until all four positive EI values for each flow group of direct runoff and baseflow are obtained for better accuracy both in direct runoff and baseflow.