Behavioural modelling - the process

The purpose of the FastStats Modelling Environment is to discover behavioural features which distinguish an analysis group from a base group, and then to use these features to create and evaluate models.

Likely stages in the process:

Select Behavioural Modelling type Fixed or Event Driven
Create an Analysis and a Base Group
Generate Behavioural Features
Build (Train ) the modelling environment - e.g. 01/01/2022
1. Use the dimensions view and Profile view to evaluate how predictive each dimension was on the training date.
2. Use the associations view to check appropriateness of using each dimension
3. Were the dimensions predictive?
  1. If YES - go to step 5
  2. If NO - return to step 1
Select an Evaluation date, possibly more – e.g. 01/01/2023 and 01/01/2024
1. compare the data on the training date with that of the evaluation date using the profile tab
2. Was the evaluation successful/ did it show the same predictive power as the results from the training date?
  1. If YES - go to step 6
  2. If NO - return to step 1
Further evaluation of the model in the Model Report tool.
1. Create a VV with a fixed date which you wish to evaluate as the Scoring Date – e.g. 01/06/2024. Does the model make a powerful prediction?
  1. If YES - create a VV for the date you wish to produce the model
  2. If NO - return to step 1

The following diagram provides a visual representation of the process for creating behaviour-based Models: