Predictive modeling

 

 

Background: Predictive modelling is used in many industries like banking, Retail, Telecom, Pharma, etc. and it is used to model an outcome/event like customer attrition.

 

Approach: Use the historical data to identify the significant relationships between a future outcome & many other correlated features in the past. We follow a set process to identify good features & create new features, data sampling, model building & validations.

 

Tools: SQL, R/Python

 

Techniques: Traditional parametric techniques like logistic regression, Machine learning algorithms like Random forest, Neural networks, etc can be applied.

 

Implementation: A simple mathematical formula can be coded in SQL or a Python script to be placed on a cloud as an automated scoring process

 

Examples: Credit card fraud detection, email spam detection, customer attrition are good examples for predictive modelling.

 

Customer attrition model: