Classification Prediction of Breast Cancer Based on Machine Learning

Table 2

Summary of filtering methods in feature selection.

Name	Variable type	Selection rules and variable requirements
Missing percentage	Univariate	Eliminate features that have too many missing samples and are difficult to fill
Variance		Exclude features with variance close to or equal to 0 apply to categorical variables
Frequency		Eliminate features that are overly concentrated on a certain category of values
Pearson’s correlation coefficient	Multivariate	Features with correlation coefficients close to or equal to 0 are removed, but the sample needs to follow a normal distribution
Analysis of variance		Exclude features with an F value that is too low, or features with a value <0.05. And the population sample is required to have homogeneity of variance and independence between samples
Kendall tau rank correlation coefficient		Exclude features with correlation coefficients close to or equal to 0, and require the categories to be ordered
Mutual information		Eliminate features with mutual information close to or equal to 0