Abstract:

According to the defects that smoothing coefficient in traditional cubic ES prediction model is fixed, the traditional one cannot track time series due to the change over time, and can not reflect the influence of historical data on the prediction results over different time periods, an improved cubic ES method is presented in this paper, which is based on the principle of minimum  sum of squared errors and blanket search algorithm to obtain the smothering factor of dynamic adjustment. It is greatly adaptable to the data with a large range of fluctuations and nonlinear regular change. Through the analysis on the simulation case of railway passenger traffic volume, the self-adaptive and dynamic cubic ES method can better adapt to the trend of time series with high prediction accuracy.

Key words: cubic exponential smoothing method, time series, smoothing coefficient, sum of squared errors, traffic prediction