3. Model Building Routine

Refer to the “Pre-requisites” prior to proceeding with the next section.

This is the first step to be executed

This chapter contains the following sections:

3.1 Contextual Information

This is the first time the Machine Learning is being setup for Oracle FLEXCUBE.

As part of this process, historical FLEXCUBE data will be referenced and requisite data transformation routine will be called. Customer attrition model is the only one that will be built. The customer life time value and segmentation do not have a persistent model object and gets executed runtime.

3.2 Model Building

User(s)

It is suggested that only the FLEXCUBE user with admin rights should have access to the model building screens

Launch ‘Model Execution and Control’ screen from the menu

Machine Learning Retail -> Model Administration -> Model Execution and Control

.

Select "Model Build" radio button

Data Fields

Remarks

Period Start Date

Models get built on historical data. This is the historical period start date to be considered

Period End Date

Models get built on historical data. This is the historical period end date to be considered

Attrition Date

This field represents the future prediction window. E.g. if we want to have a prediction window of 3 months, this would be Period End Date + 3 months

Silent Period to consider for attrition identification

While closed customers are considered as churned, for open customers having sufficiently long silent period of no transaction can also be considered as churned. This is the field that determines the number of silent days, to be considered for open customers to be tagged as churned.

Suggestions:

 

Process ‘Model Build’ routine

Click ‘process’ button to trigger model building

Note