Abstract:

There is a growing tendency to embed the entire data mining process within a single software solution. These data mining suites have data capture, pre-processing, model building and deployment all integrated. In two articles the authors will discuss the merits of interactive model building. In this first paper the `why' question will be answered; in the second paper, the authors will demonstrate `how' to do this. This `interactivity' basically consists of overriding statistical parameters as they are derived from training data. The authors propose interactive model building as an alternative to automatic model building. First, interactive model building generates more knowledge on customer behaviour and on the structure of the data. Secondly, deliberately influencing the way models are being generated actually leads to better predictions of customer behaviour. The authors illustrate how the context of the business problem can and should be taken into account when developing models.Journal of Targeting, Measurement and Analysis for Marketing (2007) 15, 137-145. doi:10.1057/palgrave.jt.5750045

Document Type: Research article

DOI: 10.1057/palgrave.jt.5750045