Knowledge Discovery using Least Squares Support Vector Machine Classifiers: a Direct Marketing Case

The case involves the detection and qualification of the most relevant predictors for repeat-purchase modelling in a direct marketing setting. Analysis is based on a wrapped form of feature selection using a sensitivity based pruning heuristic to guide a greedy, step-wise and backward traversal of the input space. For this purpose, we make use of a powerful and promising least squares version (LS-SVM) for support vector machine classification. The set-up is based upon the standard R(ecency) F(requency) M(onetary) modelling semantics. Results indicate that elimination of redundant/irrelevant features allows to significantly reduce model complexity. The empirical findings also highlight the importance of Frequency and Monetary variables, whilst the Recency variable category seems to be of lesser importance. Results also point to the added value of including non-RFM variables for improving customer profiling.