Python3 贝叶斯分类

 # -*- coding: utf-8 -*-
 """
 Created on Tue Jan 16 20:11:07 2018

 @author: markli
 """
 import numpy as np;
 import pandas as pd;

 """
 贝叶斯分类
 训练数据格式
 X = [x1,x2,...xm]; m*n
 xi = [xi1,xi2,...xin].T
 Y = [y1,y2,...ym];1*m
 """

 class Bayes:
     def __init__(self,lamda,region):
         """
         lamda 贝叶斯修正参数
         region 特征属性取值域，类标签取值域
         例如两个特征取值范围分别为，A1=[1,2,3],A2=['S','M','L']
         类标签取值：C=[1,-1]，region=[A1,A2,C]
         """
         self.lamda = lamda;
         #存放类标签域
         self.Y = region[-1];
         #存放特征取值域
         self.X = region[:-1];

         #存放先验概率 P(Y = Ck)
         self.PrioPro = np.zeros((1,len(region[-1])));

         #存放条件概率 P(Xj = ajl | Y = Ck)
         self.ConditionalPro = [];
         for i in range(len(region)-1):
             cp = np.zeros((len(region[-1]),len(region[i])));
             self.ConditionalPro.append(cp);

     def fit(self,TrainData):
         """
         计算先验概率和条件概率，建立模型
         TrainData 为二维数组
         TrainData 列的顺序与region中特征属性顺序一致
         TrainData 最后一列为类别
         """
         N = len(TrainData);
         K = len(self.Y);
         TrainData = TrainData.astype(np.str);

         NumofCk = pd.value_counts(TrainData[:,-1], sort=True); #Series 类型
         CountOfCk = [NumofCk[ck] for ck in self.Y]; #list类型
         self.PrioPro = [(ck+self.lamda) / (N + K * self.lamda) for ck in CountOfCk];

         j=0;
         for ck in self.Y:
             #选出类别为Ck的数据
             DataofCk = TrainData[np.where(TrainData[:,-1]==ck)];
             n = len(DataofCk);
             #选出第i个特征的数据
             for i in range(len(self.X)):
                 DataofCkandXi = DataofCk[:,i];
                 Numofaj = pd.value_counts(DataofCkandXi,sort=True); #为第i个特征的每个特征值计数
                 Countofaj = [Numofaj[aj] for aj in self.X[i]];
                 S = len(self.X[i]);
                 self.ConditionalPro[i][j] = [(aj+self.lamda) / (n+S * self.lamda) for aj in Countofaj];
             j = j+1;

     def predict(self,TestData):
         """
         预测实例，为其分类
         测试数据没有类别列，其余数据格式与训练数据格式一致
         """
         predictY = [];
         for i in range(len(TestData)):
             x = TestData[i];
             y = self.GetLable(x);
             predictY.append(y);
         return predictY;

     def GetLable(self,x):
         """
         输入一个测试实例x ，输出使其后验概率最大的类别y
         """
         pro = [];
         n = len(x);
         for j in range(len(self.Y)):
             p = 1;
             for i in range(n):
                 feature = self.ConditionalPro[i];
                 fi = self.X[i] #获得第i个特征的值域
                 index = fi.index(x[i]);
                 p = p * feature[j][index];

             p = p * self.PrioPro[j];
             pro.append(p);

         y = self.Y[np.argmax(pro)];

         return y;

测试代码

 # -*- coding: utf-8 -*-
 """
 Created on Wed Jan 17 15:29:14 2018

 @author: markli
 """

 import numpy as np;
 import pandas as pd;
 from BayesClass import Bayes

 A1 = ['','',''];
 A2 = ['S','M','L'];
 C = ['','-1'];

 Data = np.array([[1,'S',-1],[1,'M',-1],[1,'M',1],[1,'S',1],[1,'S',-1],
                  [2,'S',-1],[2,'M',-1],[2,'M',1],[2,'L',1],[2,'L',1],
                  [3,'L',1],[3,'M',1],[3,'M',1],[3,'L',1],[3,'L',-1]]);
 test = np.array([[2,'S']]);
 B = Bayes(0,[A1,A2,C]);
 B.fit(Data);
 y = B.predict(test);
 print(y);