RE: [Classifier4j-devel] Simple Implementation

[Nick L. <nl...@es...>]

Yes, it's possible to use a flat text file. You'll need to write your own
implementation of IWordsDataSource to do it.

Here's some pseudo code to do the training:

            IWordsDataSource wds = new SimpleWordsDataSource();
            ITrainableClassifier classifier = new BayesianClassifier(wds);

		for each message in blackMsgs 
			classifier.teachNonMatch(message)

		for each message in whiteMsgs 
			classifier.teachMatch(message)

		for each message in bayMsgs 
			int result = classifier.classify(message)

Nick

-----Original Message-----
From: Kashif [mailto:ks...@ai...]
Sent: Tuesday, 13 July 2004 12:45 PM
To: cla...@li...
Subject: [Classifier4j-devel] Simple Implementation


Hi
 
I am doing a research on Bayesian filters. I am trying to implement
classifier 4J and will appreciate a bit of help.
 
Please note that I am doing a very basic implementation, without using a
JDBC Connection. Later I might move on to JDBC and MySQL.
 
Background:
I am using searchterm and arrays for my blacklist, whitelist and baylist (ie
potential emails for Bayesian Filtering).
 
            SearchTerm blackSt = new OrTerm(blackListSearch);
            SearchTerm whiteSt = new OrTerm(whiteListSearch);
            SearchTerm baySt = new NotTerm(new OrTerm(blackListSearch));
//  If not in BlackList
    
            Message[ ] blackMsgs = folder.search(blackSt);
            Message[ ] whiteMsgs = folder.search(whiteSt);
            Message[ ] bayMsgs = folder.search(baySt);       // These are
the messages which I want to filter with bayesian
    
    System.out.println("No of messages found in whitelist : " +
whiteMsgs.length);
    System.out.println("No of messages found in blacklist : " +
blackMsgs.length);
    System.out.println();
    System.out.println("No of messages ready for bayesian filter : " +
bayMsgs.length);
 
            //Implementation of Bayesian Classifier 4J
            
            IWordsDataSource wds = new SimpleWordsDataSource();
            IClassifier classifier = new BayesianClassifier(wds);
            
            System.out.println("Matches = " + classifier.classify("This is a
sentence") );
    
 
Here's the problem:
1)       I understand that I have to train the filter, need to know how I
can do it.
2)       Is it possible to use a flat file (ie text file) rather than jdbc
connection