informs ny



Call Sam Koslowsky at  (212) 520-3259 or email him.



Incremental Response Modeling: Uses, Algorithms and Comparisons


presentation slides


Suppose a retailer sends 20% off coupons to customers to encourage them to shop. Who should receive coupons?

A traditional response model will appear good in this situation, because it successfully predicts those who will shop. Unfortunately, many of these people would have shopped even without the coupon, so the net effect of sending them a coupon is to reduce margin. Incremental Response Modeling (IRM) seeks to predict those who will shop only if they receive a coupon.

This talk will introduce Incremental Response Modeling, and present several algorithms for building IRM predictive models. We will also look at initial results in an ongoing comparison of methods using actual customer data from retail and financial domains.



Steve Gallant
Director of Data Mining Services for KXEN

Steve Gallant helps customers solve analytic problems. He has consulted for major banks, insurance companies, and retailers. His research has centered around modeling algorithms, neural networks, and information retrieval. Dr. Gallant is the author of over 40 articles and conference papers, a book on Neural Network Learning, and three patents.

He has a Ph.D. in Operations Research from Stanford University, and was Associate Professor of Computer Science at Northeastern University.



Wednesday May 12, 2010 Noon - 2 PM


The Penn Club, 30 West 44th Street, New York, NY



$50 for non-members, $40 for chapter members,
$5 surcharge for walk-in without reservation.



Call Sam Koslowsky at  (212) 520-3259 or email him.

Browse the meeting schedule , or get reservation information .