bayesianspamchecker.java

Practical AI in Java 人工智能编程

import java.util.*;

public class BayesianSpamChecker {
    public BayesianSpamChecker(Hashtable wordProbabilities) {
        this.wordProbabilities = wordProbabilities;
    }

    public float probabilityOfSpam(String text) throws Exception {
        Vector words = Tokenizer.getTokens(text);
        double probs = 1, probs_minus_one = 1;
        HashSet set = new HashSet();
        for (int i = 0; i < 15; i++) {
            interesting[i] = -1;
            wrds[i] = "";
        }
        // get unique words:
        for (int i = 0, size = words.size(); i < size; i++) {
            set.add(words.get(i));
        }
        Iterator iter = set.iterator();
        while (iter.hasNext()) {
            String w = (String) iter.next();
            float f = 0.5f;
            Float ff = (Float) wordProbabilities.get(w);
            if (ff != null) {
                f = ff.floatValue();
            }
            float interest = Math.abs(f - 0.5f);
            // see if this word is more interesting than on of the 15:
            // first, find the smallest value to replace (if it is smaller than the
            // value in 'interest':
            float minVal = 9999;
            int minIndex = 0;
            for (int i = 0; i < 15; i++) {
                if (minVal > interesting[i]) {
                    minVal = interesting[i];
                    minIndex = i;
                }
            }
            //System.out.println("w="+w+", minVal="+minVal+", minIndex="+minIndex);
            if (interest > minVal) {
                interesting[minIndex] = interest;
                wrds[minIndex] = w;
            }
        }
        for (int i = 0; i < 15; i++) {
            if (interesting[i] < 0) continue;
            Float ff = (Float) wordProbabilities.get(wrds[i]);
            if (ff != null) {
                float f = ff.floatValue();
                //System.out.println("word: " + wrds[i] + " interest: " + interesting[i] + " spam prob: " + f);
                probs *= f;
                probs_minus_one *= (1 - f);
            }
        }

        //if (Math.abs(f - 0.5f) > 0.05f) {
        //    probs *= f;
        //    probs_minus_one *= (1 - f);
        //}
        return (float) (probs / (probs + probs_minus_one));
    }

    private Hashtable wordProbabilities;

    float[] interesting = new float[15];
    String[] wrds = new String[15];
}