|
Activity Number:
|
17
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 3, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #302089 |
|
Title:
|
Predictive Modeling of Bank Portfolio Data on Home Equity Products
|
|
Author(s):
|
Oksana V. Shcherbak*+ and Duane L. Steffey
|
|
Companies:
|
San Diego State University and San Diego State University
|
|
Address:
|
8880 Rio San Diego Dr. #1045, San Diego, CA, 92108,
|
|
Keywords:
|
stochastic gradient boosting ; predictive modeling ; scoring function ; multiple additive regression trees ; nonlinearn nonparametric regression
|
|
Abstract:
|
The paper will discuss modeling of home equity potential using data of existing bank customers. The goals are to develop a score function that will help to target those bank customers who are presently not home equity customers but who are very likely to accept a home equity offer and utilize the offered product; to test the score function by comparing the response rates of a targeted group versus a non-targeted group; to explore the possibility of applying the model criteria to noncustomer data in order to identify potential new customers for the bank and for these types of products. TreeNet technology, a nonlinear nonparametric regression tool based on stochastic gradient boosting, is considered in the study. This new modeling approach to machine learning implements stagewise function approximation where each stage uses model residuals from the previous stage and incorporates a new, disjoint subsample of the training data. The final model is a collection of weighted and summed trees combined via averaging.
|
