Executing What If Analysis 

Use

You can use the What If Analysis to predict prospective customer behavior based on specific attributes of that customer. This is applicable in situations where online predictions are required. For example, when a customer applies for a loan, you can predict if the customer is creditworthy or not, based on certain attributes such as income, profession or transaction history.

What If Analysis can be used for the data mining methods Clustering and Decision Tree Analysis.

Prerequisites

·        The model for the analysis needs to be fully created and activated (see Creating, Changing, and Activating a Model)

·        The model must be trained (see Creating Analysis Process for Training)

Procedure

...

       1.      Position the cursor on the relevant data mining model, use the right-hand mouse to choose What If Analysis in the context menu. Alternatively, you can double-click on the relevant model and choose the What If option from the tool bar.

This invokes the What If Analysis screen.

       2.      Enter the model values for the prediction fields.

The model field values for prediction must be selected from the given set of values. If you enter a value that is not present in the training data, a warning message will be generated and this value will not be considered for prediction.

       3.      Choose Predict to view the results.

Result

In a decision tree, the prediction result is displayed as a histogram in both tabular and graphical form. In the histogram, all possible class outcomes are shown along with their probability and support in the training data. For example, in the case of a customer applying for a loan, the what-if prediction shows the probability of the customer getting accepted or rejected.

In clustering, there is no graphical representation of the What If Analysis. The result contains the first and the second best cluster under which the predicted case falls.