Research Of Survival Analysis Based On LASSO-CD Methd

As a statistical method of studying censored data, survival analysis has been applied to biomedicine. In recent years, its application has been expanded to a great deal of fields such as industry, commerce and social sciences in which there exist problems like anticipation and estimation for given incidents.This article proposes the survival analysis model studying method in high dimensional censored data environment on the basis of existing survival analysis model.1. This paper expands the application area of traditional Cox proportional hazards model. Traditional Cox proportional hazards model is always applied in a certain low dimensional space, while its flexibility and accuracy would be affected when it be applied in high dimensional space,. To solve this problem, this article proposes a Cox proportional hazards model variable selecting method in high dimensional censored data environment by using nonlinear mathematic ability of The Least Absolute Shrinkage and Select Operator (LASSO).2. This article comparatively profoundly studies LASSO method, Coordinate Descent arithmetic and Cox proportional hazards model, on the basis of which, this article optimizes traditional Cox proportional hazards model to apply to LASSO method. The optimizing model is operated with variable selection by Coordinate Descent arithmetic in following step. As a result, the optimizing model variable subset is selected by principle of AIC and BIC truncation.3. This article applies Matlab combining with the Monte-Carlo simulation method to produce increasing Type-Ⅱ randomly censored high dimensional data which are suitable for survival analysis. The operating condition of model variable selecting method for Cox proportional hazards is studied in different sample size and different censored ratio. The study is proved to be effective by numerical simulation and to a real PBC data set, it is proved that the method in this article has a promising application.