Purchase prediction in database marketing with the probrough system

Abstract

We describe how probabilistic rough classifiers, generated by the rule induction system ProbRough, were used for purchase prediction and discovering knowledge on customer behavior patterns. The decision rules were induced from the mail-order company database. Construction of ProbRough is based on the idea of the attribute space partition and was inspired by the rough set theory. The system’s beam search strategy in a space of models is guided by the global cost criterion. The system accepts noisy and inconsistent data with missing attribute values. Background knowledge is used in the form of prior probabilities of decisions and different costs of misclassification. ProbRough provided a lot of useful information about the problem of customer response modeling, and demonstrated its usefulness and efficiency as a data mining tool.