PCA-ELMBJ算法下右删失生存数据分析

Analysis of right censored survival data based on PCA-ELMBJ algorithm

针对右删失不完全数据,通过Buckley-James估计器(BJ)进行无偏变换.同时,将主成分分析(PCA)方法与算法相结合,提出了PCA-ELMBJ集成算法.通过随机生成不同删失率下的模拟数据,采用C-index评价指标,验证了PCA-ELMBJ集成算法比原先的ELMBJ集成算法预测效果更好,运行时间更短.最后,通过对带有右删失的TCGAmirna和EMTAB386基因特征数据的实证分析,进一步说明了所提出的PCA-ELMBJ集成算法的准确性和有效性.

An unbiased transformation is conducted by the Buckley-James estimator for the right-censored incomplete data.Meanwhile,the PCA(principal component analysis)method is combined with the ELMBJ algorithm to propose the PCA-ELMBJ ensemble algorithm.By randomly generating simulated data at different censoring proportions and using C-index,it is verified that the PCA-ELMBJ ensemble algorithm has better prediction performance and shorter running time than the original ELMBJ ensemble algorithm.Finally,through the empirical analysis of TCGAmirna and EMTAB386 gene expression data,the paper further demonstrates the accuracy and effectiveness of the proposed PCA-ELMBJ ensemble algorithm.