A novel forecasting method for the random time series

作者：Xiaohui, Hu^*; Lvjun, Zhan; Yun, Xue; Weixing, Zhou; Guixi, Liu

来源：International Journal of Digital Content Technology and Its Applications, 2012, 6(12): 223-231.

DOI：10.4156/jdcta.vol6.issue12.27

• Changing regularity
• Development trends
• Forecasting methods
• Microscopic variation
• Non-linear regression
• Probability matrixes
• Random time series
• Transition probabilities

A novel forecasting method for the random time series has been proposed in this paper. This paper considers the random time series as the superposition of the development trend curve and the volatility around the trend curve, and then uses the linear regression or the nonlinear regression or grey GM (1, 1) model to fit the trend curves; further applies the markov probability matrix forecasting method to analysis the changing regularity of the data's fluctuating up and down around the trend curve. The prediction method can make full use of given information of the observed data, use a tendency curve to reflect the overall development regularity of the system, and divide the tendency curve into several dynamic states, further put the markov transition probability in use to characterize the microscopic variation of the system, therefore provide a new analysis approach and forecasting method for the prediction of the random time data sequence. Compared with the common prediction methods, this method has the characteristics of simple computation and high precision.

• 出版日期2012-7
• 单位华南师范大学; 华南理工大学