Feature Selection For Semi-parametric Mixture Cure Model

In clinical medicine research,there are often cured individuals in experimental data.In cancer clinical research,the same cluster of individuals are usually related.In this paper,a semi-parametric mixture cure model is used to analyze the cluster survival data with cure part.Among them,the cure rate was fitted by logistic regression function,and the distribution of the survival function of the untreated individuals obeyed the proportional risk model.In this paper,we use the penalty generalized estimation equation(PGEE)method to select variables.By adding SCAD penalty to the generalized estimation equation,we compress the parameter estimation value,so as to achieve the effect of parameter estimation and variable selection at the same time.In this paper,BIC criterion is used to determine the tuning parameter ?,ES algorithm is used to solve the parameter estimation,and bootstrap method is used to estimate the variance of the parameters.Through numerical simulation,it is found that the method proposed in this paper can effectively select variables and greatly reduce the amount of calculation.Finally,the proposed method is applied to tonsil cancer data.