|
Activity Number:
|
306
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #300489 |
|
Title:
|
Random Forests for Multiclass Classification: Random Multinomial Logit Applied to an ACRM Problem
|
|
Author(s):
|
Anita Prinzie*+ and Dirk Van den Poel+
|
|
Companies:
|
The University of Manchester and Ghent University
|
|
Address:
|
MBS West - Booth Street West, Manchester, M15 6PB, United Kingdom Department of Marketing, Gent, B-9000, Belgium
|
|
Keywords:
|
analytical customer relationship management ; random forests ; multinomial logit ; feature selection ; random multinomial logit ; RMNL
|
|
Abstract:
|
Multinomial Logit (MNL) is considered to be the standard in multi-class classification and is commonly applied within the analytical CRM (Customer Relationship Management) domain. Unfortunately, to date, it is unable to handle huge feature spaces typical of CRM applications. Conversely, random forests, unlike MNL easily handles high-dimensional feature spaces. This paper investigates the potential of applying the random forests principles to the MNL framework. We propose the Random Multinomial Logit (RMNL) (i.e., a random forest of MNLs) and compare its performance to that of (a) MNL, (b) random forests. We illustrate the RMNL on a cross-sell CRM problem within the home-appliances industry. The results indicate a substantial increase in performance of the RMNL model to that of the MNL model with expert feature selection. (See www.crm.UGent.be.)
|
