readme.txt

bayes network classifier toolbox 贝叶斯网络分类器工具箱的bin文件

--------------------------------------------------
jBNC - Bayesian Network Classifier Toolkit in Java
--------------------------------------------------

jBNC is a Java toolkit for training, testing, and applying
Bayesian Network Classifiers. Implemented classifiers have been
shown to perform well in a variety of artificial intelligence,
machine learning, and data mining applications.

jBNC is primarily intended as a library for creation of Bayesian
Classifier networks. Several algorithms for creation of networks
are included. 

To aid testing the quality of classifier network a couple of simple 
command line tools for training and testing are included, see section 
TOOLS for more details.

There is also a separate package called jBNC-WEKA that integrates 
jBNC with WEKA (Waikato Environment for Knowledge Analysis 
http://www.cs.waikato.ac.nz/~ml). jBNC-WEKA allows creation of 
jBNC classifiers from within WEKA, in particular, using WEKA's 
graphical user interface. For more info see jBNC homepage.

Author  : Jarek Sacha
e-mail  : [e-mail]
homepage: http://jbnc.sourceforge.net

-------------------------------
 INSTALATION OF BINARY PACKAGE 
-------------------------------
Uncompress content of jBNC_bin_ver.zip to a directory of choice. Add 
jBNC.jar to your system CLASSPATH variable. Alternatively, location
of jBNC.jar can be specified as a direct parameter to a Java command.
For instance, you can specify, classpath as parameter to Java:

  java -cp jBNC.jar jbnc.Classifier
  
The above command sets classpath to 'jBNC.jar' while executing 
jbnc.Classifier.

-------
 TOOLS 
-------
Beside a library for creation of Bayesian Networks, jBNC provides a couple
of simple command line tools to help testing created Bayesian Classifier 
Networks. Tools can be run from a command line typing:

  java jbnc.Tool_Name options

where Tool_Name is Classifier, CrossVal, DatasetInfo, or ReVal. If options
are not specified, a tool will print help describing options available for 
that tool.

Classifier  - Loads training data set, creates Bayesian Network Classifier
              and tests it on a test data set. Type of classifier and its
	      parameters are specified through command line options.
	     
CrossVal    - Similar to Classifier tool, but performs cross validation 
              test.
	     
DatasetInfo - Prints information about a data set.

ReVal       - Change values of discrete (nominal) attributes to integers

------------
 CHANGE LOG
------------

v.1.2.2
-------

Bug fix release: Classes were compiled to Java version 1.5 preventing them from
running on earlier Java versions. Fixed build.xml to produce classes compatible
with Java 1.3 and newer.

v.1.2.1
-------

* Serialization added to some classes to support use of jBNC in WEKA 3.4.2.

* Support for serialization also required some changes to JavaBayes. Updated
  JavaBayes source is available from jBNC web site and CVS.
  
* Ant build script modified to first automatically build updated JavaBayes.

* Test cases moved to a separate directory 'test/src'. Naming convention
  changed to postfix.

v.1.2
-----
* Class for entropy-based discretization of continuous features 
  (jbnc.discretize.EntropyDiscretizer).
  
* Improved build.xml. Added support for unit testing and reporting.

v.1.1
-----
* Refactorings to simplify WEKA wrappings (see project jBNC-WEKA)


----------------------
 BUILDING FROM SOURCE
----------------------

jBNC can be build from source using Ant (http://ant.apache.org). jBNC depends
on JavaBayes source build. You can yet updated JavaBayes source at jBNC web site
either by downloading JavaBayes source package 
  http://sourceforge.net/project/showfiles.php?group_id=49871&package_id=120700
or by checking JavaBayes from jBNC CVS 
  http://sourceforge.net/cvs/?group_id=49871
  
JavaBayes and jBNC source directories should be in parallel.

To build jBNC.jar type 'ant jar' in the jBNC's top source directory 
(where jBNC's 'build.xml' file is located). The Ant's build script will first
build JavaBayes then jBNC itself.

To get description of additional build targets in the 'build.xml' type 
'ant -projecthelp'.  

-------
 LINKS
-------
applet.JavaBayes - http://www.cs.cmu.edu/~javabayes
WEKA      - http://www.cs.waikato.ac.nz/~ml
Ant       - http://ant.apache.org