Merits of interactive decision tree building — Part 2: How to do it

Abstract

In the previous paper, the authors explained why there is a tendency to embed data mining solutions in end-to-end software solutions. The advantages of integrated data mining solutions lie in making the process less people dependent, but the disadvantages are that learning from the mining process is hampered. The topic of the previous paper was why to build data mining models interactively. In this paper, the authors will explain how to build decision trees interactively. In this paper, we will demonstrate how interactive model building generates more knowledge on customer behaviour and on the structure of the data. The authors present guidelines for interactive tree building. These guidelines demonstrate how knowledge on when and how the model will be deployed can be taken into account to optimise the model. Furthermore, they illustrate how the context of the business problem that is being addressed with data mining can and should be taken into account when developing models.