项目作者: darrencl

项目描述 :
Lightweight feature selection package in Julia
高级语言: Julia
项目地址: git://github.com/darrencl/FeatureSelectors.jl.git
创建时间: 2020-02-27T04:31:02Z
项目社区:https://github.com/darrencl/FeatureSelectors.jl

开源协议:MIT License

下载


FeatureSelectors.jl

Stable
Dev
CI
Codecov

Simple tool to select feature based on the statistical relationship between features to target variable. The currently implemented feature is based on:

  • Correlation
  • P-value, which can be obtained by either Chi-square or F test

Quick start

  1. julia> using RDatasets, FeatureSelectors, DataFrames
  2. julia> boston = dataset("MASS", "Boston");
  3. julia> selector = UnivariateFeatureSelector(method=pearson_correlation, k=5)
  4. UnivariateFeatureSelector(FeatureSelectors.pearson_correlation, 5, nothing)
  5. julia> select_features(
  6. selector,
  7. boston[:, Not(:MedV)],
  8. boston.MedV
  9. )
  10. 5-element Vector{String}:
  11. "LStat"
  12. "Rm"
  13. "PTRatio"
  14. "Indus"
  15. "Tax"