李航《统计学习方法》朴素贝叶斯分类代码实现及解析

参考博客贝叶斯估计（python 版） 

import numpy as np

train_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_data = np.array([2, "S"])

fea = [{"1": 0, "2": 0, "3": 0}, {"S": 0, "M": 0, "L": 0}]
Y = {"-1": 0, "1": 0}

fea_concat_y = {}

train_set_num = len(train_data)#训练集样本数目15


for i in range(train_set_num):#遍历每一个样本
    data = train_data[i]
    size = len(data)#size=3
    '''
    key = x0,y or x1,y
    '''
    for j in range(size - 1):#j=0,1
        fea[j][str(data[j])] = fea[j][str(data[j])] + 1#对每一特征维度中的特征进行累加计数
        # key为统计所有（x1,1),(x1,-1),(x2,1),(x2,-1)的个数，
        # 目的是计算P（X=x,y=1)orP(X=x,y=-1)类似概率
        key = data[j] + ',' + data[size - 1]
        if key not in fea_concat_y:
            fea_concat_y[key] = 1
        else:
            fea_concat_y[key] = fea_concat_y[key] + 1
    # 对Y = {"-1": 0, "1": 0}中的分类-1和1进行计数，目的是求P（y=1）和P（y=-1）
    Y[data[size - 1]] = Y[data[size - 1]] + 1


'''
特征个数 len(test_data)
'''

p_result = .0
predict_label = ''
for y in Y:
    p_y = Y[y] * 1.0 / train_set_num#P(Y=y)
    p_mul = p_y
    for j in range(len(test_data)):
        key = test_data[j] + ',' + y
        #计算P(X=x1|Y=y)，即在Y=y的条件下X=x1的概率
        p_x_1_y = fea_concat_y[key] * 1.0 / train_set_num / p_y
        p_mul = p_mul * p_x_1_y#进行累乘
    if p_mul > p_result:
        p_result = p_mul
        predict_label = y
print("test case is:", test_data)
print("predict label is:", predict_label)
print("prob is", p_result)