Friday, November 6, 2009

ADAPA 2.18 released. New features and enhanced PMML support!



Zementis is constantly adding new features to ADAPA. A great example is the release of ADAPA version 2.18 (released on November 4, 2009). This release is packed with new features for its web services component as well as enhanced PMML (Predictive Model Markup Language) support.





Web S
ervices

A new web service has been added that brings the following functionality:
  • Import or export models

  • Remove models

  • Describe models (get model information)

  • Apply multiple models over multiple records

This new service comes as a second option besides the existing "model as an operation" web services. As of this version, the "model as an operation" web service is referred to as the RPC Web Service, reflecting the fact that every model is practically turned into a separate Remote Procedure Call. Note that in order to accommodate the new web service, the address (URL) and namespaces of the existing ones have been modified. Any applications developed against the web services of ADAPA 2.17 (or earlier) will require minor modifications to work with 2.18. Details can be found in the Web Services documentation which is accessible from the ADAPA Console Help Page.

For a list of all the Web Services available for Predictive Models in ADAPA, please click HERE.









ADAPA 2.18 adds support for three additional aspects of PMML:
  • DefineFunction in TransformationDictionary: A DefineFunction in PMML allows for a function to be defined once and called multiple times with a different list of arguments.

  • Targets for classification models (Targets were already supported for regression models)

  • TwoStep Clustering Models: A TwoStep Clustering Model is a clustering model exported by SPSS which is able to cluster very large datasets with mixed continuous and discrete data types. An overview of this model is given in the SPSS Enabling Technologies Division web page. Essentially, this method calculates distances between points using the distance measure of J. D. Banfield and A. E. Raftery ("Model-based Gaussian and non-Gaussian clustering",Biometrics,49,1993,pp. 803-821) as extended by M. Meila and D. Heckerman ("An experimental comparision of several clustering methods", Microsoft Research Technical Report,1998).

    For tips on how to represent TwoStep Clustering Models in PMML or for how to use the DefineFunction and Targets elements, please refer to the ADAPA Predictive Analytics Guide available for download from the ADAPA Console Help page.

    For more information on PMML support in ADAPA, please click HERE.

No comments:






Copyright © 2009-2014 Zementis Incorporated. All rights reserved.

Privacy - Terms Of Use - Contact Us