CaptainFest / Future_selection

Different future selection methods

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Future_selection

Data:

  • 99 objects with 10000 features train dataset
  • 99 objects with 10000 features test dataset

    Classificator

  • Random forest from sklearn library

    Used future selection methods:

  • Filters:
    1. feature_importances_ method from sklearn library
    2. euclidean distance (data must be normalized)
    3. coefficient correlation
  • Wrapper:
    1. Forward selection

    Results

    1. Score without future selection: 0.7373737373737373
    2. Score with feature_importances_: 0.7474747474747475
    3. Score with euclidean distance: 0.8080808080808081
    4. Score with coefficient correlation: 0.8282828282828283
    5. Score with wrapper: 0.7676767676767676

    About

    Different future selection methods

    License:MIT License


    Languages

    Language:Jupyter Notebook 100.0%