Feature Engineering

• emsemble
  • Kaggle Ensembling Guide | MLWave
  • Stacking
  • Blending
  • Stacknet
    • h2oai/pystacknet

Preprocessing

Data cleaning

• Get rid of the corresponding instance
• Get rid of the whole attribute.
• interpolate the missing value
  • Set the values to some value (zero, the mean, the median, etc.).

df.dropna(subset=["attr1"]) # option 1

df.drop("attr1", axis=1) # option 2

median = df["attr1"].median() # option 3
df["attr1"].fillna(median, inplace=True)

Reference

• Best Practices for Feature Engineering