自回归建模预测及其分析

PREDICTION BY THE MODELLING OF SELF-REGRESSIVE METHOD AND ITS ANALYSIS

自回归建模(简称AR)是一个变量的未来值用它的历史数据加权之和来表示。本文用三种时序列(均值、一阶差分、二阶差分)进行AR建模预测,择优录用最佳序列。文中对乾县、乌加河、易县等台站的倾斜仪资料进行了自回归建模预测分析,用AIC信息准则判定最佳模型阶数,其预测精度与[4]文方法相当,并有提高。文中对预测时段长度进行了探讨。用确定的预测限对1985年10月5日沙河驿5.3级地震作了预测检验,证实震前有明显的异常。

The modelling of self-regressive method (abbr. AR)means that future values of a variable are expressed by adding weight-sum of historical values of this variable. In this paper, the AR-predictions use three kinds of time-subsequence (average value, first order-difference, second order-difference) and afterwards an optimum time-subsequence can be chosen for the application. The data of tiltmeters at Qian-Xian, Wujiahe, Yixian station have been analysed by using AR-prediction. The optimum order of model is determined by AIC information-criterion. The prediction accuracy is equivalent to the accuracy of [4] paper and is higher than [4]. In this paper, the length of prediction time-section is discussed. According to determinate prediction-limit, the 1985 Shaheyi earthquake of M_s=5.3 is subjected to a prediction test, and the obvious anomalies before this earthquake are also confirmed.