The Comparison Of Markov Chain Point Predicting Methods And Time Series Analysis Predicting Methods Based On Statistical Test

The article introduced the basic theory and method of ARIMA model which is a popular method in time series analysis, then made simple introduction to the procession of data analysis and predicting with Decision time software. The focus of this article is three Markov chains point predicting method presented by the author, and the three methods are based on the Markov chains state predicting. At last, the author made a comparison of two kinds of predicting methods, and researched some factors influenced predicting accuracy. The consequences are listed below:1. Mean predicting accuracy of Markov methods is better than time series analysis method, and the predicting accuracy of ADMCPP method is the lowest in three Markov methods, there is no big difference between SPMCPP method and WMCPP method.2. Point predicting methods of Markov methods are suitable for different distribution.3. Historical data length(C) awfully influenced predicting accuracy, and Markov method demands are less than time series analysis methods.4. C_v and C_s have prominent influence on predicting accuracy, it descends when C_v increased and ascends when C_s increased.5. One step self-correlation coefficient has little influence on predicting accuracy.6. In Markov predicting methods, the state classing has influence on predicting accuracy.7. In Markov predicting methods, weight of previous time value have some influence on predicting accuracy.The valuable point of the article is given three point predicting methods of Markov chains, and compared with time series analysis predicting methods, and find out that Markov methods are better.