Diabetes key characteristic parameter acquisition method

Abstract

The invention discloses a diabetes key feature parameter acquisition method, which comprises the following steps of: performing algebraic combination and standardization processing on diabetes related original data to obtain a diabetes risk factor candidate feature set; the method comprises the following steps: screening features related to diabetes through an RReliefF algorithm, and constructing a maximum correlation feature set; redundant features irrelevant to diabetes mellitus are removed through an mRMR algorithm, and a maximum-correlation and minimum-redundancy feature set is constructed; and performing causal replacement by adopting an improved FCL causal discovery method to obtain a diabetes key feature set. By means of algebraic combination, the complexity of the diabetes feature set can be greatly increased, and the diabetes features with higher prediction contribution degree can be selected conveniently; a maximum-correlation minimum-redundancy diabetes feature set is obtained by using an RReliefF algorithm and an mRMR algorithm, and the dimensionality of the diabetes feature set is reduced; an improved FCL algorithm is used to carry out causal replacement before diabetes features, and a diabetes feature set with better contribution degree is obtained.

Description

technical field [0001] The invention relates to the technical field of feature extraction, in particular to a method for obtaining key feature parameters of diabetes. Background technique [0002] Diabetes has become an epidemic disease that seriously affects human health. How to effectively prevent and treat diabetes has become an urgent problem to be solved. Analyzing the relationship between risk factors and diabetes and establishing a diabetes prediction model is the key to revealing the pathogenesis of diabetes, and it is also an effective auxiliary means for diabetes prevention and treatment. The existing diabetes prediction methods can be mainly divided into two categories: regression statistics and intelligent prediction. [0003] (1) Diabetes prediction method based on regression statistics. There are mainly Logistic, Cox, risk factor scoring, ROC and other methods. Among them, the Logistic method has the advantages of simple sample requirements and convenient mo...

AI Technical Summary

Problems solved by technology

[0005] Regression statistical prediction methods are usually aimed at specific groups of people and are greatly affected by samples, so the adaptability of the method is poor; intelligent prediction methods are easy to deal with high-dimensional data and nonlinear problems, and can improve the prediction classification accuracy to varying degrees, and the prediction accuracy is affected by the quality of the data set. restrict

Images