What will happen?

Forming Analytical Questions

Konstantinos Kattidis

Data Analytics Lead

Understanding what will happen

Predictive analytics is focused on:

Business question:

How much should we order to meet customer demand and minimize inventory costs?

Analytical question:

What is the predicted customer demand for product x based on various factors?

Technique:

Sales forecast, line chart

Business question:

How can we identify fraudulent claims and mitigate the risk of fraudulent payouts?

Analytical question:

What is the likelihood of a claim being fraudulent based on various variables from historical claim data?

Technique:

Workflow to a classification model

Business question:

How can we improve patient outcomes by identifying which patients are at risk for certain health conditions?

Analytical question:

Can we predict which patients are at risk for certain health conditions based on relevant variables?

Technique:

Text predictive modeling workflow for health diagnosis

Business question:

How can we identify which loan applications are most likely to result in default?

Analytical questions:

Descriptive analytics: Which loan applications resulted in default in the past?

Diagnostic analytics: Why certain loan applications are usually resulting in defaults?

Predictive analytics: Can we predict the likelihood of a loan application resulting in default based on various variables in the application?

Technique:

Classification prediction algorithm such as logistic regression
The model can be applied to new loan applications to predict the likelihood of default
This would help mitigate the risk of default and improve overall portfolio performance

Workflow to a classification model for loan default predictions

Forming Analytical Questions