摘要
文章基于贝叶斯后验推理及遗传算法研究了线性回归模型多结构变点的变点检测方法,其中变点的数量和位置均未知。首先引入参数的无信息先验并基于后验分布得到变点信息的贝叶斯后验概率;然后基于后验概率定义一个施瓦兹-贝叶斯信息准则并应用遗传算法快速得到变点的个数及其位置,通过数值模拟验证了新方法的有效性和计算的快速性;最后,将所提方法应用于气象时序数据和股票交易价格两个实际数据的多变点检测问题,得到了有意义的结果。
This paper investigates the detection of multi-structural change points in linear regression model based on the Bayesian posterior inference and genetic algorithm, where the number and location of change points are unknown. Firstly, the no-information priors of parameters are introduced, and Bayesian posterior probability of change point information is obtained based on posteriori distribution. Then, a Schwartz-Bayes information criterion is defined based on posterior probability, and the number and location of change points are obtained quickly by using genetic algorithm. The effectiveness of the new method and the speed of calculation are verified by numerical simulation. Finally, the proposed method is applied to the change point detection of meteorological time series data and stock price data, with the significant results obtained.
作者
胡丹青
赵为华
Hu Danqing;Zhao Weihua(School of Sciences,Nantong University,Nantong Jiangsu 226019,China)
出处
《统计与决策》
CSSCI
北大核心
2022年第6期21-25,共5页
Statistics & Decision
基金
国家级大学生创新训练计划项目(202110304005Z)。
