       |
|
News Letter : May 2005 |
BAYESIALAB 3.3
|
Main new features available in BayesiaLab 3.3
New Data Importation Wizard
In its 3.3 release, BayesiaLab integrates a new data importation wizard that allows a better memory management, that improves the data importation processing, that comes with a greater flexibility in use, and the offers a set of new functionalities:
- Sampling: possibility to import a sample of the database:
- random sample by setting a size
- random sample by setting a percentage
- custom sample by setting indices
- Discretization: new interactive interface to discretize the continuous variables using the following methods:
- decision tree
- equal distances
- equal frequencies
- manually, by designing the intervals on the repartition function or by modifying the result of one of the three automatic methods
- Discretization type selection:
- possibility to have different discretization methods for each continuous variable
- possibility to apply a manually defined discretization on a selected set of variables
- Aggregation: possibility to aggregate discrete modalities :
- discrete modality aggregation by directly selecting the modalities
- possibility to invoke a wizard that uses the variable correlations with respect to a discrete target variable
- Weighting variable : possibility to set a variable as a line weighting variable
Enrichment of the analysis and learning tools
BayesiaLab 3.3 also integrates new analysis functionalities:
- Monitor sorting: monitors can be sorted with respect to the target variable correlations. It is then possible to quickly get the probabilistic profile of the target variable
- Addition of the positive and negative maximum variations in the target analysis report. These measures indicate the modalities that have been the most impacted by the observation of the corresponding target modality. The first one indicates the modality with the greatest increase whereas the second one indicates the modality with the greatest decrease
- Node comments integration in the analysis reports to improve the readability of the results
- Arc Trimming Interface: possibility to display/hide the arcs according to their KL force. This function allows to focus the analysis on the most important relations. This is particularly useful for network of great size
- Suppression of the virtually disconnected nodes : possibility to remove the nodes isolated by the use of Arc Trimming Interface. It allows then to delete the nodes that do not have any probabilistic relations with an importance threshold greater than the one defined by the trimming interface
- Probabilistic relations Threshold: possibility to modify the weight of the structural complexity in the score used by the learning algorithms. It allows then to modify the minimal threshold of the strength of the probabilistic relations. This new functionality, particularly useful for data file with few lines, is available in the Preferences menu item.
- Temporal indices: The file of temporal indices allows associating temporal
indices to nodes. These temporal indices are taken into account by the BayesiaLab's structural learning
algorithms to prevent adding arcs from future to past nodes
Increased ergonomics
Taking care of its ease of use, BayesiaLab 3.3 improves again its ergonomics:
- Dictionaries: possibility to use text file to associate to each node a comment and/or a colour category in order to increase the Bayesian networks legibility
- Graph Rotation: new graphical tools to rotate your Bayesian networks in order to find the best layout
- Fast and easy access to the last used text files, databases, or Bayesian networks
- Improvement of the data table entry
- Addition of the "Close all" function
- Quick layout algorithm: possibility to specify in the Preferences the layout algorithm associated with the shortcut key (P).
User guide
The online BayesiaLab's help file is now available in a user guide.
Download it by clicking below :
 |
Do not wait more to test these new features,
|
Bayesia also proposes to you to ...
consult the two new pages on the web site :
 |
18 good reasons to use Bayesia's technologies for your marketing problems To understand why great names of the Marketing Research field have already chosen BayesiaLab |
 |
8 good reasons to use Bayesia's technologies for the risk management To understand why Bayesian networks are essential for reliability and risk management problems |
have a look at the two new dynamic presentations that illustrate the BayesiaLab's use on two real applications:
