Improving direct mail targeting through customer response modeling

摘要: Data-mining algorithms (CHAID, CART and neural networks) perform best.Logistic regression linear discriminant analysis are statistical alternatives.Quadratic analysis, naive Bayes, C4.5 k-NN algorithm badly. Direct marketing is an important tool in the promotion mix of companies, amongst which direct mailing crucial. One approach to improve mail targeting response modeling, i.e. a predictive modeling that assigns future probabilities customers based on their history with company. The contributions literature three-fold. First, we introduce well-known data-mining classification techniques (logistic regression, quadratic networks, decision trees, including CHAID, C4.5, algorithm) community. Second, run benchmarking study using above classifiers four real-life datasets. 10-fold cross-validated area under receiver operating characteristics curve used as evaluation metric. Third, give managerial insights facilitate classifier choice trade-off between interpretability performance classifier. findings benchmark show well this test bed, followed by simplistic like logistic analysis. It shown yield poor performance.