python代码实现朴素贝叶斯分类器（离散变量）

import pandas as pd
import numpy as np

def fit(df,lambda_=0):
    class_name = df.columns.to_list()[-1]
    feature_names = df.columns.to_list()[:-1]

    # 类标记计数
    N = df.shape[0]
    class_count = df[class_name].value_counts().to_dict()

    # 先验概率，使用的是贝叶斯估计
    pri_prob = {}
    for calss_key, calss_value in class_count.items():
        pri_prob[calss_key] = (class_count[calss_key] + lambda_)/(N + len(class_count) * lambda_)

    # 特征计数和条件概率，使用的是贝叶斯估计
    feature_count = {}
    con_prob = {}
    for calss_key, calss_value in class_count.items():
        feature = {}
        feature_prob = {}
        for column in range(len(feature_names)):
            feature[feature_names[column]] = df[df[class_name] == calss_key][feature_names[column]].value_counts().to_dict()
            for feature_key, feature_value in feature.items():
                feature_value_prob = {}
                for feature_value_key, feature_value_value in feature_value.items():
                    feature_value_prob[feature_value_key] = (feature_value[feature_value_key] + lambda_) / (class_count[calss_key] + lambda_ * len(feature_value))
                feature_prob[feature_names[column]] = feature_value_prob
        feature_count[calss_key] = feature
        con_prob[calss_key] = feature_prob

    return pri_prob, con_prob


def predict(pri_prob, con_prob, dict):
    for i in range(len(dict)):
        class_prob = {}
        for class_key, class_value in pri_prob.items():
            prob_list = [pri_prob[class_key]]   # 存放先验概率和条件概率
            for data_key, data_value in dict[i].items():
                prob_list.append(con_prob[class_key][data_key][data_value])
            class_prob[class_key] = np.prod(prob_list)  # 计算概率
        dict[i]['Y'] = max(class_prob, key=class_prob.get) # 取最大概率对应的类标记
        print(i, dict[i])



if __name__=='__main__':
    df = pd.read_excel('test.xlsx')
    pri_prob, con_prob = fit(df, lambda_=1)
    df_predict = pd.read_excel('predict.xlsx').T.to_dict()
    predict(pri_prob, con_prob, df_predict)

结果如下：
 

0 {'X1': 2, 'X2': 'S', 'Y': -1}
1 {'X1': 3, 'X2': 'L', 'Y': 1}

代码的难点在计算条件概率和求解目标函数。

朴素贝叶斯分类器主要理解条件概率意义、条件独立性假设以及贝叶斯估计的作用。

训练数据（'test.xlsx'）如下（可复制粘贴到xlsx）：

X1	X2	Y
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

测试数据（'predict.xlsx'）如下（可复制粘贴到xlsx）：

X1	X2
2	S
3	L