Target group selection in retail banking through Neuro-fuzzy data mining and extensive pre- and postprocessing

Abstract

Based on a real-life problem, the target group selection for a bank’s database marketing campaign, we will examine the capacity of Neuro-Fuzzy Systems (NFS) for Data Mining. NFS promise to combine the benefits of both fuzzy systems and neural networks, and are thus able to learn IF-THEN-rules, which are easy to interpret, from data. However, they often need extensive preprocessing efforts, especially concerning the imputation of missing values and the selection of relevant attributes and cases. In this paper we will demonstrate innovative solutions for various pre- and postprocessing tasks as well as the results from the NEFCLASS Neuro-Fuzzy software package.