The purpose of this paper is to discuss and assess the structural characteristics (conceptual utility) of the most popular classification and predictive techniques employed in customer relationship management and customer scoring and to evaluate their classification and predictive precision.
A sample of customers' credit rating and socio‐demographic profiles are employed to evaluate the analytic and classification properties of discriminant analysis, binary logistic regression, artificial neural networks, C5 algorithm, and regression trees employing Chi‐squared Automatic Interaction Detector (CHAID).
With regards to interpretability and the conceptual utility of the parameters generated by the five techniques, logistic regression provides easily interpretable parameters through its logit. The logits can be interpreted in the same way as regression slopes. In addition, the logits can be converted to odds providing a common sense evaluation of the relative importance of each independent variable. Finally, the technique provides robust statistical tests to evaluate the model parameters. Finally, both CHAID and the C5 algorithm provide visual tools (regression tree) and semantic rules (rule set for classification) to facilitate the interpretation of the model parameters. These can be highly desirable properties when the researcher attempts to explain the conceptual and operational foundations of the model.
Most treatments of complex classification procedures have been undertaken idiosyncratically, that is, evaluating only one technique. This paper evaluates and compares the conceptual utility and predictive precision of five different classification techniques on a moderate sample size and provides clear guidelines in technique selection when undertaking customer scoring and classification.
