基于RJMCMC的对数正态分布均值多变点检测
Log-normal Distribution Mean Change Point Detection Based on RJMCMC
陈丽曲 1黄介武1
作者信息
- 1. 贵州民族大学数据科学与信息工程学院,贵阳 550025
- 折叠
摘要
针对数据呈现偏态分布且存在变点的情况,建立对数正态分布均值多变点模型,给出该分布模型的似然函数,利用可逆跳马尔可夫链蒙特卡罗(RJMCMC)算法来计算多个变化点的数量和位置的后验概率.设计了四种类型的跳跃,并给出了每种跳跃的接受概率.仿真研究表明,基于RJMCMC的方法能够有效地检测对数正态分布序列中均值参数的变化个数及变点位置.最后,通过实证分析说明了所建立的模型是切实可行的.
Abstract
For situations that the data exhibits skewed distributions and contains change points,a multiple-change point model for the mean of the log-normal distribution is established.The likelihood function of this distribution model is provided,and the reversible jump Markov chain Monte Carlo(RJMCMC)algorithm is utilized to compute the posterior probability of the number and locations of multiple change points.Four types of jumps are designed,along with the acceptance probabilities for each type.Simulation studies indicate that the RJMCMC-based method can effectively detect the number of changes and the locations of change points of the mean parameter within log-normal distribution sequences.Finally,empirical analysis demonstrates the practical feasibility of the established model.
关键词
对数正态分布/贝叶斯方法/RJMCMC方法/均值变点Key words
log-normal distribution/bayesian methods/RJMCMC method/mean change points引用本文复制引用
出版年
2024
NETL
NSTL
万方数据