Publications

Trimmed Bagging

Bagging has been found to be successful in increasing the predictive performance of unstable classifiers. Bagging draws bootstrap samples from the training sample, applies the classifier to each bootstrap sample, and then averages overal lobtained classification rules. The idea of trimmed bagging is to exclude the bootstrapped classification rules that yield the highest error rates, as estimated by the out-of-bag error rate, and to aggregate over the remaining ones. In this note we explore th...

Computational Statistics and Data Analysis, 52 (1), 362-368.
Croux, C., Joossens, K. and Lemmens, A.
2007

Bagging and Boosting Classification Trees to Predict Churn

In this article, the authors explore the bagging and boosting classification techniques. They apply the two techniques to a customer database of an anonymous U.S. wireless telecommunications company, and both significantly improve accuracy in predicting churn. This higher predictive performance could ultimately lead to incremental profits for companies that use these methods. Furthermore, the results recommend the use of a balanced sampling scheme when predicting a rare event from large data ...

Journal of Marketing Research, 43(2), 276-286.
Lemmens, A. and Croux, C.
2006

On the Predictive Content of Production Surveys: a Pan-European Study

For over 40 years, Business Tendency Surveys have been collected in multiple member states of the European Union. Previous research has studied the predictive content of the expectation variables included in those surveys through bivariate, within-country, Granger-causality tests. These tests have resulted in mixed conclusions. We extend previous research in various ways, as we (i) explicitly allow for cross-country influences, and (ii) do so using both bivariate and multivariate Granger-caus...

International Journal of Forecasting, 21(2), 363-375.
Lemmens, A., Croux, C. and Dekimpe, M.G.
2005