In this work we use ROC (Receiver Operating Characteristic) analysis to evaluate customer subgroups detected by different machine learning approaches in a marketing database. A direct mailing model with a marginal cost per mailing and an average expected profit per new customer has been assumed. In order to identify optimal mailing strategies for different marketing situations, we introduce the normalised profit curve, which extends the ROC curve by not only identifying the optimal subgroup in a given context, but also indicating the expected profit. In this sense, the analysis presents a link between data mining and decision support.
|Translated title of the contribution||Subgroup evaluation and decision support for a direct mailing marketing problem|
|Title of host publication||Unknown|
|Editors||Christophe Giraud-Carrier, Nada Lavrac, Steve Moyle|
|Publisher||ECML/PKDD'01 workshop notes|
|Pages||45 - 56|
|Number of pages||11|
|Publication status||Published - Sep 2001|