• Why to ignore a Feature:
  • it's noisy
  • it's causing overfitting
  • it's highly correlated with an other features
  • speed up the training/testing process

There are two big univariate feature selection tools in sklearn:

• SelectKBest selects the K features that ate most powerful
• SelectPercentile select the X% of features that are most powerful

We can also apply Recursive Feature Elimination