摘要
对于方差相等且已知的正态分布序列的均值多变点问题,一般采用二分法,把问题简化为仅有唯一变点问题。本文提出利用贝叶斯统计的方法识别唯一变点的位置,主要是利用贝叶斯统计的方法求出变点位置的后验概率密度,然后利用贝叶斯信息准则给出变点有无的检验准则,并在有变点的情况下给出变点位置的具体估计。
Dichotomy is usually used to deal with problems of mean change-point models for normal distribution sequence.The question is simplified as one-variant question.In this paper,Bayesian statistical method is proposed to identify the location of the only change point.The method mainly uses Bayesian statistics to find the posterior probability density of the change point location,and then uses the Bayesian information criterion to give the test criteria for change point,and gives the specific location of the change point.
作者
郭卫娟
GUO Wei-juan(School of Mathematics&Economics,Institute of Large Data Modeling and Intelligent Computing,Hubei University of Education,Wuhan 430205,China)
出处
《湖北第二师范学院学报》
2020年第8期8-11,共4页
Journal of Hubei University of Education
基金
湖北第二师范学院大数据建模与智能计算研究所项目“变点位置识别的贝叶斯机械学习方法”。
关键词
正态分布序列
二分法
贝叶斯后验密度
贝叶斯信息准则
normal distribution sequence
dichotomy
Bayesian posterior density
Bayesian Information Criterion
