Study On The Applications Of Random Lasso In Logistic Model

In the current information age of large data,the data information obtained is too much and too complicated.At the same time,more and more variables are collected,it is difficult to extract valuable variables from it.Therefore,it is very important to process the model through effective variable selection methods.The generalized linear model is a very widely used model in practice.There have been many studies and applications of variable selection in this model,however the variable selection methods dealing with high dimensional and strongly correlated data are rarely studied in this field.In this paper,the Random Lasso method is extended to a generalized linear model for variable selection,and applied in the field of disease diagnosis and treatment.Through the establishment of Logistic classification model for a set of disease diagnosis and treatment data,the disease classification can be judged.This paper also compares the Random Lasso method with the common variable selection methods,such as Lasso,Elastic Net and Adaptive Lasso methods,to evaluate the model classification performance under each method.By plotting the ROC curve and calculating the AUC value of the area under the curve,and combining with other classification evaluation indicators,it is concluded that the model classification effect established by the Random Lasso method is better than the other three common variable selection methods.