General information

Brick provides an assessment of the predictive power of independent variables with respect to the binary dependent variable via Weight of Evidence (WoE) and Information Value (IV) calculation:

• WoE - reflects the strength of each category in predicting the desired value of the target variable. It is equal to the natural logarithm of the division of the part of all non-events that belong to the category's sub-sampling and the part of events correspondingly:

  $$ WoE_{category_i} = \ln\Bigg( \frac{(non\events{category_i})/(non\events{total})}{(events_{category_i})/(events_{total})}\Bigg) $$

In the case of continuous variable data is preliminary split into 10 parts - 10 intervals with the equal density.

• IV - a measure of the predictive power of the independent variable based on the separate categories' WoE values:

$$ IV_{category_i} = \sum\bigg( \frac{non\events{category_i}} {non\events{total}}-\frac{events_{category_i}}{events_{total}}\bigg)\times WoE_{cetegory_i} $$

Information Value explanation

Description

Brick Location

Bricks → Analytics → Data Insights → Information Value

Bricks → Analytics → Credit Scoring → Information Value

Bricks → Use Cases → Credit Scoring → Features Engineering → Information Value

Brick Parameters

• Target

  A binary variable that is used as a target variable in a binary classification problem. The information value of the separate predictors is calculated with respect to the specified target. The target variable should present in the input dataset and takes two values - (0, 1).

• Columns

  List of possible columns for selection. It is possible to choose several columns for filtering by clicking on the '+' button in the brick settings and specify the way of their processing:

  • remove all mentioned columns from dataset and proceed with rest ones as with predictors
  • use the selected columns as predictors and proceed with them

• Remove all except selected

  The binary flag, which determines the behaviour in the context of the selected columns

• Method

  The property, which determines the view of the final results representation:

  • Information Value with WoE - detailed information about predictors that includes the information value of the variable, information value of its separate categories, and their weight of evidence scores
  • Information Value only - the short representation of the Information Value brick results, which includes the listing of the selected predictors with their information values with respect to the specified target variable

Brick Inputs/Outputs

• Inputs

  Brick takes the dataset, which contains the binary target variable and independent predictors