Sampling Tricks to improve your Accuracy in Credit Scoring and Customer Reationship Management
Published 17 May 2008
Sampling Tricks to improve your Accuracy in Credit Scoring and Customer Reationship Management
Keynote speech at SAS M2008 Data Mining Conference, Las Vegas, USA
Dr. Sven F. Crone presented a session keynote on at the SAS M2008 Data Mining conference, the largest Data Mining conference in the world. His presentation outlined new findings on Data Mining corporate datasets in credit scoring, direct marketing and customer relationship management with imbalanced class distributions.
The research presented under the title “The presentation “A Bag of Tricks for Your Balancing Act: How to Increase Predictive Accuracy on Imbalanced Datasets” was developed jointly with Dr. Steve Finlay and Dr. Didier Soopramanien, both Lancaster Centre for Forecasting. It shows, that a simple oversampling of the minority class (e.g. those defaulting on a loan or reacting to a mass mailing) often leads to increased accuracy across most established data mining methods.
For additional information please contact Dr. Sven F. Crone, deputy director of the Lancaster Centre for Forecasting.
