Hierarchical important feature selection method based on clinical high-dimensional breast cancer data

Abstract

The invention discloses a hierarchical important feature selection method based on clinical high-dimensional breast cancer data. The feature selection method comprises steps of statistical feature selection and integrated feature selection. At the statistical feature selection step, a single factor analysis method is employed; and features having obvious impacts on outcome variables are selected preliminarily by different statistics and checking. At the integrated feature selection step, a gradient lifting tree is established; after model training, a feature importance score is obtained; witha designed and verified importance score threshold, selection of features having obvious impacts on outcome variables is realized. Therefore, problems of too high data feature dimension, excessive redundant features and data disorder during the clinical breast cancer prediction modeling process are solved effectively; the redundant or meaningless features in high-dimensional clinical breast cancerdata can be excluded; and thus a few of features having the important impact on the breast cancer modeling are selected and the high accuracy and practicability of the breast cancer model are ensured.

Description

technical field [0001] The invention relates to the fields of computer technology, statistical machine learning technology, feature engineering technology and the like. Background technique [0002] Breast cancer is the malignant tumor with the highest incidence rate among women in the world, which seriously threatens women's health. Breast cancer patients are usually intervened through surgery, chemotherapy and other treatment measures, and may face the risk of recurrence at any time after treatment. Scientifically assessing and predicting the survival status of breast cancer patients can assist doctors in formulating appropriate treatment plans, providing new support for reducing the risk of recurrence and improving prognosis. [0003] To realize the assessment and prediction of the survival status of breast cancer patients, such as the recurrence-free survival rate, a machine learning prediction model can be established based on breast cancer clinical data. However, cli...

AI Technical Summary

Problems solved by technology

[0013] The purpose of the present invention is to solve the problem of too high clinical data dimension in establishing breast cancer survival prediction model

Use the hierarchical feature selection method combining statistical feature selection and integrated feature selection to solve the problem of important feature extraction and model practicability

Images