西瓜书7.3 朴素贝叶斯分类器

实现拉普拉斯修正的朴素贝叶斯分类器

定义输入

输入为Object二维数组objects，objects[i][j]表示第i个西瓜的第j个属性

 Object[][]objects={ {"青绿", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.697, 0.460, "好瓜"},
                {"乌黑", "蜷缩", "沉闷", "清晰", "凹陷", "硬滑", 0.774, 0.376, "好瓜"},
                {"乌黑", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.634, 0.264, "好瓜"},
                {"青绿", "蜷缩", "沉闷", "清晰", "凹陷", "硬滑", 0.608, 0.318, "好瓜"},
                {"浅白", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.556, 0.215, "好瓜"},
                {"青绿", "稍蜷", "浊响", "清晰", "稍凹", "软粘", 0.403, 0.237, "好瓜"},
                {"乌黑", "稍蜷", "浊响", "稍糊", "稍凹", "软粘", 0.481, 0.149, "好瓜"},
                {"乌黑", "稍蜷", "浊响", "清晰", "稍凹", "硬滑", 0.437, 0.211, "好瓜"},
                {"乌黑", "稍蜷", "沉闷", "稍糊", "稍凹", "硬滑", 0.666, 0.091, "坏瓜"},
                {"青绿", "硬挺", "清脆", "清晰", "平坦", "软粘", 0.243, 0.267, "坏瓜"},
                {"浅白", "硬挺", "清脆", "模糊", "平坦", "硬滑", 0.245, 0.057, "坏瓜"},
                {"浅白", "蜷缩", "浊响", "模糊", "平坦", "软粘", 0.343, 0.099, "坏瓜"},
                {"青绿", "稍蜷", "浊响", "稍糊", "凹陷", "硬滑", 0.639, 0.161, "坏瓜"},
                {"浅白", "稍蜷", "沉闷", "稍糊", "凹陷", "硬滑", 0.657, 0.198, "坏瓜"},
                {"乌黑", "稍蜷", "浊响", "清晰", "稍凹", "软粘", 0.360, 0.370, "坏瓜"},
                {"浅白", "蜷缩", "浊响", "模糊", "平坦", "硬滑", 0.593, 0.042, "坏瓜"},
                {"青绿", "蜷缩", "沉闷", "稍糊", "稍凹", "硬滑", 0.719, 0.103, "坏瓜"}
        };

为了方便，将每个objects[]封装为一个类，属性名与属性值放入一个map中

public class Entity {
    private Map<String,Object> map=new HashMap<String, Object>();
    //种类名
    private String classifyLabel;

    public Entity(Object[] objects, String[] labels) {
        for(int i=0;i<objects.length;i++){
            map.put(labels[i],objects[i]);
        }
        classifyLabel=labels[labels.length-1];
    }

代码

Main

package com.fly;

public class Main {
    public static void main(String[] args) {
        String[]labels={"色泽","根蒂","敲声","纹理","脐部","触感","密度","含糖率","好瓜"};
        Object[][]objects={ {"青绿", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.697, 0.460, "好瓜"},
                {"乌黑", "蜷缩", "沉闷", "清晰", "凹陷", "硬滑", 0.774, 0.376, "好瓜"},
                {"乌黑", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.634, 0.264, "好瓜"},
                {"青绿", "蜷缩", "沉闷", "清晰", "凹陷", "硬滑", 0.608, 0.318, "好瓜"},
                {"浅白", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.556, 0.215, "好瓜"},
                {"青绿", "稍蜷", "浊响", "清晰", "稍凹", "软粘", 0.403, 0.237, "好瓜"},
                {"乌黑", "稍蜷", "浊响", "稍糊", "稍凹", "软粘", 0.481, 0.149, "好瓜"},
                {"乌黑", "稍蜷", "浊响", "清晰", "稍凹", "硬滑", 0.437, 0.211, "好瓜"},
                {"乌黑", "稍蜷", "沉闷", "稍糊", "稍凹", "硬滑", 0.666, 0.091, "坏瓜"},
                {"青绿", "硬挺", "清脆", "清晰", "平坦", "软粘", 0.243, 0.267, "坏瓜"},
                {"浅白", "硬挺", "清脆", "模糊", "平坦", "硬滑", 0.245, 0.057, "坏瓜"},
                {"浅白", "蜷缩", "浊响", "模糊", "平坦", "软粘", 0.343, 0.099, "坏瓜"},
                {"青绿", "稍蜷", "浊响", "稍糊", "凹陷", "硬滑", 0.639, 0.161, "坏瓜"},
                {"浅白", "稍蜷", "沉闷", "稍糊", "凹陷", "硬滑", 0.657, 0.198, "坏瓜"},
                {"乌黑", "稍蜷", "浊响", "清晰", "稍凹", "软粘", 0.360, 0.370, "坏瓜"},
                {"浅白", "蜷缩", "浊响", "模糊", "平坦", "硬滑", 0.593, 0.042, "坏瓜"},
                {"青绿", "蜷缩", "沉闷", "稍糊", "稍凹", "硬滑", 0.719, 0.103, "坏瓜"}
        };

        Object[]object=  {"青绿", "蜷缩", "浊响", "清晰", "凹陷", "硬滑", 0.697, 0.460, "？"};
        BayesClassifier classifier=new BayesClassifier(objects,labels);
        Entity entity = new Entity(object, labels);
        System.out.println("分类结果为"+classifier.getClassify(entity));
    }

}

Entity

package com.fly;

import java.util.HashMap;
import java.util.Hashtable;
import java.util.Map;
import java.util.Set;

public class Entity {
    private Map<String,Object> map=new HashMap<String, Object>();
    private String classifyLabel;

    public Entity(Object[] objects, String[] labels) {
        for(int i=0;i<objects.length;i++){
            map.put(labels[i],objects[i]);
        }
        classifyLabel=labels[labels.length-1];
    }
    //获取所有属性
    public Set<String>getProperty(){
       return map.keySet();
   }
    //获取某个属性的值
   public Object getValue(String property){
       return map.get(property);
   }

    @Override
    public String toString() {
        return "Entity{" +
                "map=" + map +
                '}';
    }


}

分类器

package com.fly;

import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import java.util.Set;

public class BayesClassifier {
    //训练集
    List<Entity>entities;
    //类名
    String classifyLabel;
    //分类结果的集合
    Set<Object>classes=new HashSet<Object>();
    public BayesClassifier(Object[][] a,String[] labels){
        entities=new ArrayList<Entity>();
        for(Object[] objects:a){
            Entity entity=new Entity(objects,labels);
            entities.add(entity);
            classes.add(objects[objects.length-1]);
        }
        classifyLabel=labels[labels.length-1];

    }

    //获取分类
    public Object getClassify(Entity entity){
        Set<String> property = entity.getProperty();
        Object classify=null;
        double p=0;
        for(Object c:classes){
            System.out.println("分类为"+c+"的概率是"+p(entity, c));
            if(p(entity,c)>p){
                p=p(entity,c);
                classify=c;
            }
        }
        return classify;
    }

    //分类为c的概率
    public double p(Entity entity,Object c){
        Set<String> properties = entity.getProperty();
        double ans=priorP(c);
        for(String property:properties){
            if(!property.equals(classifyLabel)){
                ans*=p(property,entity.getValue(property),classifyLabel,c);
            }
        }
        return ans;
    }

    //先验概率
    public double priorP(Object c){
        int num=0;
        for(Entity entity:entities){
            if (entity.getValue(classifyLabel).equals(c)){
                num++;
            }
        }
        //return num*1.0/entities.size();
        return (num+1)*1.0/(entities.size()+classes.size());
    }

    //条件概率p(x|c)
    public double p(String label1,Object x,String label2,Object c){
        double ans=0;
        if(! (x instanceof Number)){
            double num1=0;
            double num2=0;
            for(Entity entity:entities){
                Object value1 = entity.getValue(label1);
                Object value2 = entity.getValue(label2);
                //System.out.println(value1+" "+value2);
                if(value2.equals(c)){
                    num2++;
                    if(value1.equals(x)){
                        num1++;
                    }
                }
            }
            return (num1+1)/(num2+getN(label1));
        }else{
            List<Number>list=new ArrayList<Number>();
            for(Entity entity:entities){
                Object value1 = entity.getValue(label1);
                Object value2 = entity.getValue(label2);
                if(value2.equals(c)){
                    list.add((Number) value1);
                }
            }
            double average = getAverage(list);
            double deviation = getStandardDeviation(list);
            //System.out.println(average+" "+variance);
            Double number = ((Number) x).doubleValue();
            double t=-(number-average)*(number-average)/2/deviation/deviation;

            ans=1/Math.sqrt(2*Math.PI)/deviation*Math.exp(t);
            return ans;
        }
    }

    //获取label属性值的种类数量
    public double getN(String label1) {
       Set<Object> set = new HashSet<Object>();
        for(Entity entity:entities){
            set.add(entity.getValue(label1));
        }
        return set.size();
    }

    //获取标准差
    public double getStandardDeviation(List<Number>list){
        return Math.sqrt(getVariance(list));
    }

    //获取方差
    public double getVariance(List<Number>list){
        double ans=0;
        double average = getAverage(list);
        for(Number number:list){
            double t =((number.doubleValue()-average)*(number.doubleValue()-average));
            ans+=t;
        }
        return ans/(list.size()-1);
    }

    //获取平均数
    public double getAverage(List<Number>list){
       double ans=0;
        for(Number number:list){
            ans+=number.doubleValue();
        }
        return ans/list.size();
    }
}

一些问题
这里判断属性是否为离散用的 instance of Number判断，若处理数值型的标称属性会出现问题

posted on 2020-12-17 11:10 计网好难阅读(178) 评论(0) 收藏举报