0-1序列唯一变点的贝叶斯检验方法  

The Bayesian Test Method for Unique Change-Point of 0-1 Sequence

作　　者：郭卫娟 GUO Wei-juan(School of Mathemodics and Econometrics,Hubei University of Education,Wuhan 430205,China;Institute of Big Data Modeling and Intelligent Computing,Hubei University of Education,Wuhan 430205,China)

机构地区：[1]湖北第二师范学院数学与经济学院 [2]湖北第二师范学院大数据建模与智能计算研究所

出　　处：《湖北第二师范学院学报》2019年第8期21-27,共7页Journal of Hubei University of Education

摘　　要：任何好的贝叶斯方法都与先验分布的选择有关,先验分布越合理,则后验估计更合理。对于0-1序列的变点问题,本文提出根据机械学习的概念对原有先验进行改进,得到改进后的更为合理的先验,然后利用该先验得出更合理的关于变点位置的后验估计,最后借助随机模拟的方法说明该方法的优越性。Any good Bayesian method depends on the selection of prior distribution.The more reasonable the prior distribution is,the more reasonable the posterior estimation is.This paper aims to improve the original priori by using machine learning method in 0-1 sequences.A more reasonable posterior estimate of the position of the change-point is obtained.Finally,the superiority of this method is illustrated by means of stochastic simulation.

关 键 词：0-1分布序列 贝叶斯因子 机械学习 改进先验分布 

分 类 号：O212.1[理学—概率论与数理统计]