Research Of Time Series Data Base On The State_Space

Time series analysis is an important method to deal with and analyze dynamic data which is based on Probability and Statistics. Time series can be divided into smooth ones and unsmooth ones according to statistical character. In practice most of the series which we meet, especially the series reflecting social and economic phenomena are often unsmooth. Forecasting such series properly can control and supervise the development of society and economy. And then it will be far-reachingly meaningful to assort these series into parts of season, circulation, trendline and random. So forecasting and modeling of dynamic data are very important in reality and applied research.Based on general methods of build moldeling This paper mainly study state-space modeling methods based on general modeling and introduce EM algorithm, Kalman filtering-wave, filtering-smoothness and forecasting methods. According to Chinese CPI monthly data, the paper presents an applied procedure based on other procedures such as EM , Kalman filtering-wave, smoothness and uses it to analyze and forecast to obtain better results than the general methods of x-11-ARIMA.