基于二元分割的多变点估计

Multiple Change Point Estimation Based on Binary Segmentation

以含有多个均值变点的时间序列数据为研究对象,将最小二乘法与二元分割相结合,提出了最小二乘二元分割方法(LSBS),改进间隔较小时二元分割方法无法检测两个相邻变点的缺点.并对铁路旅客量数据进行实证分析,统计模拟结果显示,改进方法很大程度上提高了多变点估计的准确性.

This article takes time series data with multiple mean change points as the research object,and combines the least squares method and binary segmentation to propose the least squares binary segmentation method(LSBS).The method improves the disadvantage of binary segmentation method,which cannot detect two adjacent change points when the interval is small.The empirical analysis of railway passenger volume data is carried out,and statistical simulation results show that the improved method greatly improves the accuracy of multivariate point estimation.