KDnuggets : News : 2008 : n06 : item14 < PREVIOUS | NEXT >

Software

From: Jouffe Lionel
Date: 20 Mar 2008
Subject: Release of BayesiaLab 4.4, the leading Bayesian Network software for Data Mining

Bayesia announces the release of BayesiaLab 4.4, the software for efficient Data Mining with Bayesian Networks (academic licenses available from 140 Euros).

Here are some new features of this release:

  • Mosaic Analysis: powerful analysis tool allowing to get a graphical representation of the joint probability distribution of a subset of variables. By using the standardized Pearson residuals, these graphs can also be used to accurately analyze the statistical significance of the probabilistic relations between the variable states. The "Horizontal Diagram" version is particularly useful to analyze clustering results;

  • Dynamic Profile: new analysis report that uses an incremental optimization procedure to optimize the selected criterion (State Probability, Variable Mean or Probability Difference between two States). By taking into account the probabilistic relations between the profile variables, this analysis is particularly fitted to searching the best levers;

  • Tree Augmented Naive Bayes: partially predefined structure allowing to relax the strong hypothesis of conditional independence that holds in the Na´ve architecture where, knowing the target node value, each node becomes independent of the other nodes. This architecture is based on the Na´ve architecture on which a maximum spanning tree is learned. The prediction accuracy is better than the one obtained by the Na´ve architecture, and this algorithm is much quicker than the Augmented Na´ve algorithm;

  • Data Clustering: the clustering algorithm can now be tuned by using the Settings: the Drift measures the difference between the theoretical (during learning) and the true (after the imputation on the learning set) probability distributions of the clusters. The Purity is the mean of the data assignment probability in the clusters. Increasing these parameters increases the constraints and reduces the number of obtained clusters;

  • and, as usual, increased ergonomics of the interface and performance improvements.

You can consult the newsletter that describes more in details the 80 new functionalities of BayesiaLab 4.4,  download BayesiaLab 4.4 for a new 30 days evaluation period, and you can also have a look at the dynamic and static presentations that illustrate successful applications of BayesiaLab.

Bookmark using any bookmark manager!


KDnuggets : News : 2008 : n06 : item14 < PREVIOUS | NEXT >

Copyright © 2008 KDnuggets.   Subscribe to KDnuggets News!