A modeling Approach Based on Combined Short-term and Long-term Prediction

A combined prediction modeling method is proposed based on a short-term and long-term strategy. A prediction model is built for adapting to the short-term pattern changes of data by utilizing double exponential smoothing model and the Fourier residual correction. And then, a Markov model is employed for adjusting the trend during the short-term predicting through using long-term trend change information from historical data. Thus, the proposed method utilizes not only the real-time data to provide the characteristics information in the current pattern change, but also the historical data to provide the overall law information in the previous pattern changes. The experiment and comparison results show that the proposed method has higher prediction accuracy.