The exploration of statistical ensemble methods for market segmentation

Researchers in all fields are constantly looking for new methods to improve classification and prediction. In the field of marketing research, many of these techniques are used to gain insight on consumers behaviors and preferences. Popular classification and prediction methods in marketing research include cluster analysis and logistic regression respectively. There are other statistical ensemble methods used for prediction, such as bagging and boosting, which have been recently introduced, but are not widely used. Bagging and boosting algorithms have been used in many other fields and tend to significantly improve prediction. This dissertation compares the prediction results of these ensemble methods with a traditional method (logistic regression) for a marketing research study. While bagging and boosting might improve prediction in most cases, this analysis provides an example where the results of these ensemble methods are not favorable.