Untitled

Feature Engineering

• Numerical transformations
  • Data Centering
    • substract the mean of a dataset to each datapoint
      • divide it by the standard deviation
        • Z scores
          • StandardScaler
  • Min Max normalization
    • find the minimum and maximum data point in our entire data set and set each of those to 0 and 1, respectively.
      • Then the rest of the data points will transform to a number between 0 and 1, depending on its distance between the minimum and maximum number
        • We find that transformed number by taking the data point subtracting it from the minimum point, then dividing by the value of our maximum minus minimum.
          • MinMaxScaler
    • this transformation does not work well with data that has extreme outliers.
  • Binning data
    • process of taking numerical or categorical data and breaking it up into groups
  • Natural Log transformation
    • This transformation works well for right-skewed data and data with large outliers
    • just because some data is skewed, it does not mean that a log transformation is the best answer
• Encoding categorical variables
  • Categorical data: has more than one category
    • Nominal: data with no particular order or hierarchy to it
    • Ordinal: categories have order