Data prediction analysis method based on COX model and random survival forest

Abstract

The invention provides a data prediction analysis method based on a COX model and a random survival forest. The method comprises steps as follows: S1, a KM curve graph is made for observed values in a data set, and a survival function changing trend increasing with time is obtained from the KM curve graph; S2, the COX model is made for variables in the data set, and the variables are screened; S3, the observed values in the data set are divided into a test set and a training set in the proportion of 75%; S4, random survival forest regression is performed on the training set, and a random survival forest model is obtained; S5, the effect of the test set is evaluated with the random survival forest model. Compared with method in the prior art, the method has the benefits as follows: the COX model is firstly used for variable screening, the number of in-model variables can be greatly reduced, finally, the screened variables are subjected to fitting and prediction with a random survival forest model method, and the fitting and prediction effect of the model is judged according to an ROC graph; the software running speed can be greatly increased, and the model prediction effect is substantially improved.

Description

technical field [0001] The invention relates to the field of data prediction analysis, in particular to a data prediction analysis method based on COX model and random survival forest. Background technique [0002] Today's society is a society of rapid development, information circulation, advanced technology, communication between people is getting closer, life is becoming more and more convenient, a large amount of data is the product of this high-tech era. How to make good use of these data and mine valuable information from the data is the key for enterprises to win the competition. The random forest method is one of the commonly used machine learning methods for dealing with high-dimensional data. This method does not need to specify the distribution characteristics of parameters in advance, and can evaluate the predictive ability of each predictor variable on the outcome; at the same time, it uses internal cross-validation to evaluate its prediction error rate and can...

AI Technical Summary

Problems solved by technology

[0003] However, there are a lot of data in many analysis tasks. If it is directly imported, the R software will crash, and there are too many irrelevant variables that will affect the prediction effect.

Images