Software

Salford Systems CART(R) and MARS(tm) now included as Clementine Modeling
Nodes

As part of a partnership between SPSS and Salford Systems, the CART(R)
decision tree, the MARS(tm) predictive modeler, and the TreeNet(tm) gradient
boosting tool will be seamlessly available to Clementine users. Once the
plug-and-play modules have been installed, the new nodes will be visible on
the modeling palette for inclusion in any Clementine analysis stream.

CART is the decision tree used by Salford Systems to win the KDDCup 2000
data mining/web mining competition. The module, based on the original
proprietary source code developed by Breiman, Friedman, Olshen and Stone,
includes many new controls for tree growing, pruning, and shaping, and for
variable penalizing.

MARS is the automated regression and logistic regression tool developed
by Stanford University's Jerome Friedman.  MARS offers automatic
variable selection, interaction detection, nonlinear transformation,
missing value handling, anomaly detection and visualization for
regression and binary classification models.

TreeNet is Jerome Friedman's recent extension to boosting, offering
strong resistance to overfitting and data errors and a variety of new
modeling diagnostics and visual displays.

Information on the products can be found at
www.salford-systems.com and details on the Clementine integration
are provided at
www.salford-systems.com/clementineplus.html