朴素贝叶斯分类器（Python实现+详细源码原理）

Posted 2023-05-02

参考技术A

1、贝叶斯公式的本质： <u>由因到果，由果推因</u>

2、贝叶斯公式：

[图片上传中...(wps6.png-5fd624-1618488341725-0)]

1、朴素贝叶斯公式

x1，x2,...xn为特征集合，y为分类结果

朴素贝叶斯假设各个特征之间相互独立

分母相同情况下，我们只要保证分子最大

训练数据集
long,not_long,sweet,not_sweet,yellow,not_yellow,species
400,100,350,150,450,50,banana
0,300,150,150,300,0,orange
100,100,150,50,50,150,other_fruit
测试数据集
long,sweet,yellow
not_long,not_sweet,not_yellow
not_long,sweet,not_yellow
not_long,sweet,yellow
not_long,sweet,yellow
not_long,not_sweet,not_yellow
long,not_sweet,not_yellow
long,not_sweet,not_yellow
long,not_sweet,not_yellow
long,not_sweet,not_yellow
long,not_sweet,yellow
not_long,not_sweet,yellow
not_long,not_sweet,yellow
long,not_sweet,not_yellow
not_long,not_sweet,yellow

结果
特征值：[not_long, not_sweet, not_yellow]
预测结果：\'banana\': 0.003, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[not_long, sweet, not_yellow]
预测结果：\'banana\': 0.006999999999999999, \'orange\': 0.0, \'other_fruit\': 0.05625000000000001
水果类别：other_fruit
特征值：[not_long, sweet, yellow]
预测结果：\'banana\': 0.063, \'orange\': 0.15, \'other_fruit\': 0.018750000000000003
水果类别：orange
特征值：[not_long, sweet, yellow]
预测结果：\'banana\': 0.063, \'orange\': 0.15, \'other_fruit\': 0.018750000000000003
水果类别：orange
特征值：[not_long, not_sweet, not_yellow]
预测结果：\'banana\': 0.003, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[long, not_sweet, not_yellow]
预测结果：\'banana\': 0.012, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[long, not_sweet, not_yellow]
预测结果：\'banana\': 0.012, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[long, not_sweet, not_yellow]
预测结果：\'banana\': 0.012, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[long, not_sweet, not_yellow]
预测结果：\'banana\': 0.012, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[long, not_sweet, yellow]
预测结果：\'banana\': 0.108, \'orange\': 0.0, \'other_fruit\': 0.00625
水果类别：banana
特征值：[not_long, not_sweet, yellow]
预测结果：\'banana\': 0.027, \'orange\': 0.15, \'other_fruit\': 0.00625
水果类别：orange
特征值：[not_long, not_sweet, yellow]
预测结果：\'banana\': 0.027, \'orange\': 0.15, \'other_fruit\': 0.00625
水果类别：orange
特征值：[long, not_sweet, not_yellow]
预测结果：\'banana\': 0.012, \'orange\': 0.0, \'other_fruit\': 0.018750000000000003
水果类别：other_fruit
特征值：[not_long, not_sweet, yellow]
预测结果：\'banana\': 0.027, \'orange\': 0.15, \'other_fruit\': 0.00625
水果类别：orange

机器学习强基计划4-3：详解朴素贝叶斯分类原理(附例题+Python实现)

目录

• 0 写在前面
• 1 贝叶斯方法
• 2 贝叶斯风险
• 3 从例子出发
• 4 朴素贝叶斯分类
• • 4.1 核心原理
  • 4.2 拉普拉斯平滑
• 5 Python实现
• • 5.1 计算类先验概率
  • 5.2 计算类后验概率
  • 5.3 预测

0 写在前面

机器学习强基计划聚焦深度和广度，加深对机器学习模型的理解与应用。“深”在详细推导算法模型背后的数学原理；“广”在分析多个机器学习模型：决策树、支持向量机、贝叶斯与马尔科夫决策、强化学习等。