Machine Learning Cheat Sheet

gistfile1.txt

Pandas methods:
- df.info() - quick description about data and amount of non-null values.
- df['category_field'].value_counts() - shows all categories and how many rows share each category.
- df.describe() - show summary of numerical attributes.
- df.hist(bins=50, figsize=(20,15)) - plots a histogram for each numerical attribute.
- df.where(df['age'] < 100, 100, inplace=True) - where the condition is false - replaces the value with the second argument. 
- df.mask(df['age'] > 100, 100, inplace=True) - the opposite of where().
- df.corr() - returns linear correlation matrix between every pair of attributes
  cm = df.corr()
  cm['some_attribte'].sort_values(ascending=False)
  
- pandas.plotting.scatter_matrix(df[categories], figsize=(12,8)) - plots a correlation matrix of every category pair.
- df.plot(kind='scatter', x='age', y='income', alpha=0.1) - same as scatter_matrix, but for single category pair.