Research Of The Decision Support Algorithm Based On Exponential Smoothing Of Time Series

Nowadays, the research of forecasting system which is based on time series is one of the hotspots of time series' applications on financial system. Exponential smoothing, as a key branch of time series, is a typical forecasting and control model. Because the model have some merit characteristics and wide application. But there still three problems needs to be solved: First, the static exponent parameter is difficult to adapt to the changes in their own time series; Second, how to select a suitable exponent parameter; Third, the initial value is difficult to determine. A new model which is based on IDS(Improved Dichotomous Search) algorithm and improved dynamic index is designed by this paper, this new model can better solve these three problems, which is confirmed by experimental results.Firstly, the definition of exponential smoothing is presented. Then three models' characteristics are analyzed. How we choose these three models is based on these differences mainly.Secondly, the dichotomous search algorithm is introduced in this paper. Then the establishment process and utilization of the IDS algorithm are presented.Thirdly, through analyzing the deficiency of original exponential smoothing model, a new improved dynamic index exponential smoothing model which is based on adaptive weight is built. Also, the parameter is optimized by IDS algorithm. Then, we forecast and analyze some economic indicators, such as stock prices, gross national product and throughout of a port, by using a forecasting system which is based on the new exponential smoothing model which we have built.Finally, experimental results demonstrate that the smoothing effect is better by using the above new model and the result is obviously superior to traditional model, and the accuracy of prediction is improved.