Software

The winners of the PAKDD-2007 data mining and predictive modeling competition, Lei Chai and Mingjun Wei, used TreeNet to construct their award-winning model.

Second place (1st runner up) Salford Systems also used TreeNet to secure another in a growing list of TreeNet-based modeling successes. Based in Hangzhou, the winners who have been Salford Systems partners since 2005, offer analytical services in the Shanghai-Hangzhou corridor. The two Salford related teams worked independently, without knowledge of the others' participation, and made use of the TreeNet outputs in substantially different ways.

The PAKDD-2007 Competition required development of a cross-sell model for a financial services company offering both credit cards and home loans. The winning submission was based on a two-model ensemble of a TreeNet and a logistic regression was fit to features selected by TreeNet.

Salford software has been key in the winning models in a lengthy series of data mining competitions, including the KDDCup 2000 (1st place in each of two challenges), the Duke University/NCR Teradata Churn Modeling Competition (1st place in each of four separate challenges), KDDCup 2004 ("Most Accurate"), BI-Cup Latin America 2005 (1st place), and the PAKDD2006 (2nd runner up) aming others We are pleased to note that several of these competitions were won by Salford clients and partners.

More information on the PAKDD2007 competition can be found at:

http://levis.shu.edu.cn/pakdd2007/competition/finalwinner

Information on TreeNet, including Stanford University Professor Jerome Friedman's papers on the technology can be found at

http://www.salfordsystems.com/TreeNet.php

and general information for Salford is available at

http://www.salford-systems.com

What's this?