变点对状态空间模型预测的影响分析  

Effect of Change Points on Prediction of State Space Model

作　　者：张虎[1] 刘素勤[2] 

机构地区：[1]合肥工业大学数学学院,合肥230009 [2]安徽大学数学科学学院,合肥230039

出　　处：《合肥学院学报（自然科学版）》2012年第1期1-7,共7页Journal of Hefei University :Natural Sciences

基　　金：教育部科技重大专项(309017);国家公益性行业科研专项(GYHY201006055)资助

摘　　要：利用变点理论对上证指数进行了分析,得到了全涨全跌股指的变点个数及其所处位置.根据上证指数变点个数的不同,分别对三个不同时段的上证指数建立状态空间模型,通过比较预测结果,得出变点越少的状态空间模型,其预测精度越高的结论.最后在没有变点的情况下比较了ARIMA、自回归与状态空间模型的预测结果,说明了状态空间模型具有更好的预测效果.The all rises and falls of Shanghai stock index returns are analyzed by change point theory in this paper. The number and the locations of change points of the all rises and falls of returns are given. Then according to the number of change points in the Shanghai stock index, state space models for three different stages of the Shanghai stock index returns are established. It is concluded that the fewer change points in the Shanghai stock index, the better the state space model has the accuracy of prediction by comparing the prediction results. Finally it shows that the prediction effect of state space model is better than ARIMA model and autoregressive model when there is no change point.

关 键 词：状态空间模型 预测 变点 伽玛分布 股票指数 

分 类 号：O213.9[理学—概率论与数理统计]