从零开始数据科学与机器学习算法-朴素贝叶斯-07

朴素贝叶斯概念

例子：邮件分类问题：

N = (12/17)*(5/11)*(3/11)
S = (5/17)*(2/7)*(1/7)

print(N)
print(S)
# N>S 我们可以判断这是一封正常邮件

常见问题1

因为图2中 吗出现的次数是0 那么这封邮件就会被误判为正常邮件

解决如果遇到样本里面有0的情况，可以通过添加alpha进行解决。

alpha=1 统一增加1 确保不会出现无0的情况

N = (12/17)*((1/15)**4)*(3/15)
S = (5/17)*((5/11)**4)*(1/11)

print(N)
print(S)

print(S > N)

贝叶斯代码

数据切分

def split_our_data_by_class(dataset):# 按类别拆分数据 传入参数数据
    splited_data = dict() # 按字典储存
    for i in range(len(dataset)):
        vector = dataset[i]
        class_value = vector[-1]
        if (class_value not in splited_data):
            splited_data[class_value] =list()
        splited_data[class_value].append(vector)
    return splited_data

# create our dummy data

dataset = [ [0.8,2.3,0],
    [2.1,1.6,0],
    [2.0,3.6,0],
    [3.1,2.5,0],
    [3.8,4.7,0],
    [6.1,4.4,1],
    [8.6,0.3,1],
    [7.9,5.3,1],
    [9.1,2.5,1],
    [6.8,2.7,1]]

splited = split_our_data_by_class(dataset)
print(splited)

# 逐行打印
for label in splited:
    print(label)
    for row in splited[label]:
        print(row)

calculate mean and standard deviation(n-1)：计算平均值和标准差（n-1）

from math import sqrt

def calculate_the_mean(a_list_of_num): # 平均值函数
    mean = sum(a_list_of_num)/float(len(a_list_of_num)) # 求总和除以数据个数
    return mean

def calculate_the_standard_deviation(a_list_of_num): # 计算标准差
    the_mean = calculate_the_mean(a_list_of_num) # 调用平均值函数计算平均值
    the_variance = sum([(x-the_mean)**2 for x in a_list_of_num])/ float(len(a_list_of_num)-1)# {每个值减去平均值的平方除以数据的（个数-1）}结果相加的总和
    std = sqrt(the_variance) # 开方
    return std

def describe_our_data(dataset):# 调用函数
    description = [(calculate_the_mean(column),
                    calculate_the_standard_deviation(column),
                    len(column)) for column in zip(*dataset)]
    del(description[-1]) # 删除最后一列的分类标签
    return description

describe_our_data(dataset)

def describe_our_data_by_class(dataset):# 综合
    splited_data = split_our_data_by_class(dataset)
    data_description = dict()# 创建一个空字典来存储
    for class_value,rows in splited_data.items():# 通过迭代填入空字典
        data_description[class_value] = describe_our_data(rows)
    return data_description

description = describe_our_data_by_class(dataset)
for label in description:
    print(label)
    for row in description[label]:
        print(row)

description = describe_our_data_by_class(dataset)
type(description)

print(description)
print("---------")
print(description[0])
print("---------")
print(description[0][0])
print("--------")
print(description[0][0][2])

def calculate_class_probability(description,row): # 根据类别计算概率
    total_rows = sum([description[label][0][2] for label in description])
    probabilities = dict()
    for class_value,class_description in description.items():
        probabilities[class_value] = description[class_value][0][2]/float(total_rows)
        for i in range(len(class_description)):
            mean,stdev,count = class_description[i]
            probabilities[class_value] *= calculate_the_probability(row[i],mean,stdev)
    return probabilities

代码整合测试

from math import sqrt
from math import pi
from math import exp


def split_our_data_by_class(dataset):
    splited_data = dict()
    for i in range(len(dataset)):
        vector = dataset[i]
        class_value = vector[-1]
        if (class_value not in splited_data):
            splited_data[class_value] = list()
        splited_data[class_value].append(vector)
    return splited_data


def calculate_the_mean(a_list_of_num):
    mean = sum(a_list_of_num) / float(len(a_list_of_num))
    return mean


def calculate_the_standard_deviation(a_list_of_num):
    the_mean = calculate_the_mean(a_list_of_num)
    the_variance = sum([(x - the_mean) ** 2 for x in a_list_of_num]) / float(len(a_list_of_num) - 1)
    std = sqrt(the_variance)
    return std


def describe_our_data(dataset):
    description = [(calculate_the_mean(column),
                    calculate_the_standard_deviation(column),
                    len(column)) for column in zip(*dataset)]
    del (description[-1])
    return description


def describe_our_data_by_class(dataset):
    splited_data = split_our_data_by_class(dataset)
    data_description = dict()
    for class_value, rows in splited_data.items():
        data_description[class_value] = describe_our_data(rows)
    return data_description


def calculate_the_probability(x, mean, stdev):
    exponent = exp(-((x - mean) ** 2 / (2 * stdev ** 2)))
    result = (1 / (sqrt(2 * pi) * stdev)) * exponent
    return result


def calculate_class_probability(description, row):
    total_rows = sum([description[label][0][2] for label in description])
    probabilities = dict()
    for class_value, class_description in description.items():
        probabilities[class_value] = description[class_value][0][2] / float(total_rows)
        for i in range(len(class_description)):
            mean, stdev, count = class_description[i]
            probabilities[class_value] *= calculate_the_probability(row[i], mean, stdev)
    return probabilities


dataset = [[0.8, 2.3, 0],
           [2.1, 1.6, 0],
           [2.0, 3.6, 0],
           [3.1, 2.5, 0],
           [3.8, 4.7, 0],
           [6.1, 4.4, 1],
           [8.6, 0.3, 1],
           [7.9, 5.3, 1],
           [9.1, 2.5, 1],
           [6.8, 2.7, 1]]

description = describe_our_data_by_class(dataset)

probability = calculate_class_probability(description, dataset[0])

print(probability)