API Reference/Feature Selection
Feature Selection
Select informative columns in browser or Node.js machine-learning workflows with model-based, recursive, and univariate JavaScript feature selection in @kanaries/ml.
View as MarkdownFeature selection
Feature selection removes weak or redundant columns before prediction. It can reduce model cost, improve interpretability, and limit overfitting. @kanaries/ml provides model-based selectors and statistical scoring functions with a sklearn-like JavaScript API.
- Model-based and recursive selectors:
SelectFromModel,RFE, andRFECV - Univariate score functions:
chi2,fClassif,mutualInfoClassif, andmutualInfoRegression
Covariance Estimation and Graphical Lasso
Estimate empirical, shrunk, Ledoit-Wolf, OAS, and sparse inverse covariance matrices in JavaScript or TypeScript.
SelectFromModel, RFE, and RFECV
Run model-based and recursive feature selection in JavaScript or TypeScript with SelectFromModel, RFE, and RFECV from @kanaries/ml.