bayes分类算法之java实现

是基本的bayes分类算法实现, 原理比较繁琐, 如果看我的代码去理解bayes算法比较困难,这个版本是我的bayes算法实现的第一个版本,还有很大的优化的余地.如果有空我会更新.

 

 

/** * * Copyright 2009 5173.com, Inc. All rights reserved. * 5173.com PROPRIETARY/CONFIDENTIAL. Use is subject to license terms. * * Projet Name: BayesTest * File Name: Classifier.java * * Author: Weng xiaojun (wengxj@5173.com) * NeoCellServer * Create Date: 2009-11-6 * Version: 1.0 * Modification:<br><br> */ public class Classifier { /** * * key : 类别 * value : * key : 单词 * value: 出现次数 */ private static HashMap<String,HashMap<String,Integer> > typeMap = new HashMap<String,HashMap<String,Integer>>(); //总共有多少种word private static Set<String> wordSet = new HashSet<String>(); //存放每个类别中总共有多少单词 HashMap<String,Integer> sumInType = new HashMap<String,Integer>(); //训练 public void train(String[] words,String type) { if(typeMap.containsKey(type)) { setWordInfo(typeMap.get(type),words,type); } //如果没有此类别 else { setWordInfo(new HashMap<String,Integer>(),words,type); } for( String tp : typeMap.keySet()) { int sum = 0; for(String w : typeMap.get(tp).keySet()) { sum += 1 * typeMap.get(tp).get(w); } sumInType.put(tp, sum); } } /** * 计算每个单词的概率和保存单词的出现次数 * @param words */ private void setWordInfo(HashMap<String,Integer> childMap,String[] words,String type) { WordInfo wi ; int sumWord = 0 ; for(String w : words) { if(!childMap.containsKey(w)) { childMap.put(w, 1); } else { childMap.put(w, childMap.get(w)+1); } } setWordSet(words); typeMap.put(type, childMap); } private void setWordSet(String[] words) { for( String w : words) { wordSet.add(w); } } public String getClassification(String[] ws) { //存放每个 HashMap<String,HashMap<String,Double>> mp = new HashMap<String,HashMap<String,Double>>(); for( String type : typeMap.keySet()) { for(String w : ws) { if(mp.containsKey(type)) { if(mp.get(type).containsKey(w)) break; else { double p = (double)(typeMap.get(type).get(w)+ 1) / (double)( wordSet.size() + sumInType.get(type)); /* System.out.println("("+ typeMap.get(type).get(w) + "+ 1)/(" + wordSet.size() + "+ "+sumInType.get(type)+")"); System.out.println("type["+type+"] "+ w +" p = " + p); System.out.println("-------------------------------------------");*/ mp.get(type).put(w, p); } } else { HashMap<String,Double> submp = new HashMap<String,Double>(); int app = 0; if(typeMap.get(type).containsKey(w)) app = typeMap.get(type).get(w); double p = (double)(app+ 1) / (double)( wordSet.size() + sumInType.get(type)); /* System.out.println("("+ app + "+ 1)/(" + wordSet.size() + "+ "+sumInType.get(type)+")"); System.out.println("type["+type+"] "+ w +" p = " + p); System.out.println("-------------------------------------------");*/ submp.put(w, p); mp.put(type, submp); } } } HashMap<String,Integer> time = new HashMap<String,Integer>(); for(String w : ws) { if(time.containsKey(w)) { time.put(w, time.get(w)+1); } else { time.put(w, 1); } } String realType = null; double maxP = 0; for(String ty : mp.keySet()) { double p1 = 1; for(String w : mp.get(ty).keySet()) { p1 = p1 * Math.pow((mp.get(ty).get(w)),time.get(w)); } double p2 = 1d / mp.size() * p1; if (p2 > maxP) { maxP = p2; realType = ty; } } return realType; } public static void main(String[] args) { String[] s1 = {"Chinese","BeiJing","Chinese"}; String[] s2 = {"Chinese","Chinese","ShangHai"}; String[] s3 = {"Chinese","Japan","Tokyo"}; String[] s4 = {"Tokyo","Kyoto","Chinese"}; Classifier cf = new Classifier(); cf.train(s1, "Y"); cf.train(s2, "Y"); cf.train(s3, "N"); cf.train(s4, "N"); String[] s5 = {"Tokyo","Chinese","Tokyo","Japan"}; System.out.println("get type = "+cf.getClassification(s5)); } }