Optimal Binning

For feature selection/ elemination purposes, this component calculates IV(Information Values) for optimal categories of variables. This component also calculates WOE (Weight of Evidence) of categorized variables.

Step By Step Guide:

1- Initially, to run this component one should install Python Integration extensions.
2- For obtain a better Python node performance, pyarrow library should be installed.
3- Having installed pyarrow library, select serialization library as Apache Arrow under preferences. This option makes a huge difference as performance compared to Flatbuffers Column Serialization.
4- Then, specify desired IV threshold, target (label) and its bad category from dialog window. Target should be a string form to run this component.

Input ports

1. Type: Table
   Data
   Raw Data

Output ports

1. Type: Table
   Binning Table
   Binning Results
2. Type: Table
   IVs Based On Each Attribute
   Information Values For Each Features
3. Type: Table
   IVs Within Threshold
   Information Values For Over Threshold Features
4. Type: Table
   Data to Apply
   Data For Optimal Binning (Apply)