The combined data-mining Algorithm of short-term electric load forecasting is presented. Based on analyzing the characteristic of daily power loads

the most similar day recently to the forecasting-target day is selected as the benchmark day. With this way of arranging data

the data mining is used to obtain the relationship of the difference of influent factors and the load variance rate between the benchmark day and the forecasting target day

and also

the mining model algorithm is the weight combination of C4.5 and CART based on BP network. The holiday influence on loads is dealt with by the holiday justifying factor based on case reasoning. The highly accurate short-term load forecasting system is designed with above solutions. The accuracy and effectiveness of the combining algorithm proposed has been proved with the actual application.