应用RJMCMC方法识别线性模型异常点  

Application of RJMCMC Method in the Research of Outliers in the Linear Regression Model

作　　者：石永亮[1] 汪四水[1] 

机构地区：[1]苏州大学数学科学学院,江苏苏州215006

出　　处：《数学的实践与认识》2011年第17期115-120,共6页Mathematics in Practice and Theory

摘　　要：传统线性模型异常点识别方法容易发生误判:正常点被归为异常点或者异常点被归为正常点.为解决此类问题,提出了应用逆跳马尔科夫蒙特卡洛方法识别异常点的思想,同时将其应用于实际数据加以检验,识别效果明显好于传统方法.Traditional methods of outlier detection tend to be failure of justice： outliers arc classified to be non-outliers and vice versa. Therefore a new idea about outlier detec- tion is proposed, which is reversible jump Markov chain Monte Carlo method. With this method, outliers in a real dataset are detected and without injustice whice is just the problem traditional methods have to face with.

关 键 词：逆跳马尔科夫链蒙特卡洛方法 异常点 识别 

分 类 号：O211.62[理学—概率论与数理统计]