Dissertation Series - CRISP DM - Step Four Data Modelling

The modelling step involves the actual data mining stage of the project, which breaks down into a number of steps;

Select modelling technique

The problem being solved/objectives set out in the business understanding step will tend to indicate which modelling technique is suitable for the problem being tackled (The Modelling Agency, 2000 p.25), however those new to data mining may test a number of different modelling techniques.

Generate test design

Data mining accuracy is very important as false positives can cause serious issues between an organisation and its customers (Thuraisingham, 1999 p.93), for this reason a robust test plan must be put in place in order to validate the data mining solution (The Modelling Agency, 2000 p.28).

Build model

This is the sub-step where the data mining application specific development is undertaken and as with software design is iterative with a loop of develop, test and adjust (The Modelling Agency, 2000 p.28).

Assess model

In union with the test design carry out testing of the model, with the option of rolling back to the build model sub-step to factor in debugging (The Modelling Agency, 2000 p.29).