Box-Pierce Q检验的改进方法

发布时间：2018-04-17 04:03

本文选题：时间序列 + 平稳性　；参考：《统计与决策》2017年17期

【摘要】：Box-Pierce Q检验采用近似卡方分布分析时间序列的平稳性特征,其检验统计量的参数选取将影响到检验结果。文章多个Q值提取平稳性特征,在此基础上建立新的平稳性判定准则,该准则是自相关函数序列收敛的充分条件;采用欧氏函数作为平稳性特征的相似性度量,借助k-means聚类建立平稳性分类方法;该方法在平稳性分析过程中充分考虑了样本之间的关联性,避免了传统Box-Pierce Q检验对统计分布和临界表的过度依赖。实验结果表明,新方法能有效地处理海量时间序列数据,且准确率高于Q检验和ADF检验。
[Abstract]:The Box-Pierce Q test uses approximate chi-square distribution to analyze the stationary characteristics of time series, and the parameters of the test statistics will affect the test results.In this paper, several Q values are used to extract stationary features, and a new stationary criterion is established, which is a sufficient condition for the convergence of autocorrelation function sequences, and Euclidean function is used as the similarity measure of stationary features.The stationary classification method is established by means of k-means clustering, which takes full account of the correlation between samples in the stationary analysis process, and avoids the over-dependence of the traditional Box-Pierce Q test on the statistical distribution and critical table.The experimental results show that the new method can deal with massive time series data effectively, and the accuracy is higher than that of Q test and ADF test.
【作者单位】：南华大学经济管理学院;
【基金】：教育部人文社科青年项目(13YJCZH044)
【分类号】：O211.61

【相似文献】