贝叶斯网络实战（附代码）

1. 建立虚拟环境

conda create -n BayesNN38 python=3.8
conda activate BayesNN38
pip install pgmpy

2. 学生是否获得推荐信

代码如下：

from pgmpy.models import BayesianNetwork
from pgmpy.inference import VariableElimination


letter_bn=BayesianNetwork([
    ('D','G'),('I','G'),('I','S'),('G','L') # 指向关系 D->I
])

from pgmpy.factors.discrete import TabularCPD
d_cpd=TabularCPD(variable='D',variable_card=2,values=[[0.6],[0.4]])
# 变量名，变量取值个数，对应概率
i_cpd=TabularCPD(variable='I',variable_card=2,values=[[0.7],[0.3]])
g_cpd=TabularCPD(variable='G',variable_card=3,values=[[0.3,0.05,0.9,0.5],[0.4,0.25,0.08,0.3],[0.3,0.7,0.02,0.2]],
 # 行数等于变量取值，列数等于依赖变量总取值数（3,4）
evidence=['I','D'],evidence_card=[2,2])
# 变量名，变量取值个数，对应概率，依赖变量名，依赖变量取值
s_cpd=TabularCPD(variable='S',variable_card=2,values=[[0.95,0.2],[0.05,0.8]],
                 evidence=['I'],evidence_card=[2])
l_cpd=TabularCPD(variable='L',variable_card=2,values=[[0.1,0.4,0.99],[0.9,0.6,0.01]],
                 evidence=['G'],evidence_card=[3]) # evidence_card必须是列表

letter_bn.add_cpds(d_cpd,i_cpd,g_cpd,s_cpd,l_cpd)
letter_bn.check_model()  # 检查构建的模型是否合理
letter_bn.get_cpds()  # 网络中条件概率依赖关系
print(letter_bn.check_model())
print(letter_bn.get_cpds())

letter_infer=VariableElimination(letter_bn) # 变量消除
prob_I=letter_infer.query(variables=['L'],evidence={'I':1,'D':1})
print(f"prob_I:{prob_I}")

输出结果：

3. 泰坦尼克数据集预测存活人员

  这个数据集是kaggle上一个竞赛数据集，这里使用贝叶斯网络进行预测，贝叶斯网络表示如下图所示：

代码如下：

import pandas as pd
from sklearn.cluster import KMeans
from pgmpy.models import BayesianModel
from pgmpy.estimators import BayesianEstimator

'''
PassengerId => 乘客ID
Pclass => 客舱等级(1/2/3等舱位)
Name => 乘客姓名
Sex => 性别 清洗成male=1 female=0
Age => 年龄 插补后分0,1,2 代表 幼年（0-15） 成年（15-55） 老年（55-）
SibSp => 兄弟姐妹数/配偶数
Parch => 父母数/子女数
Ticket => 船票编号
Fare => 船票价格 经聚类变0 1 2 代表少 多 很多
Cabin => 客舱号 清洗成有无此项，并发现有的生存率高
Embarked => 登船港口 清洗na,填S
'''
# combine train and test set.
train = pd.read_csv('./train.csv')
test = pd.read_csv('./test.csv')
full = pd.concat([train, test], ignore_index=True)
full['Embarked'].fillna('S', inplace=True)
full.Fare.fillna(full[full.Pclass == 3]['Fare'].median(), inplace=True)
full.loc[full.Cabin.notnull(), 'Cabin'] = 1
full.loc[full.Cabin.isnull(), 'Cabin'] = 0
full.loc[full['Sex'] == 'male', 'Sex'] = 1
full.loc[full['Sex'] == 'female', 'Sex'] = 0

full['Title'] = full['Name'].apply(lambda x: x.split(',')[1].split('.')[0].strip())
nn = {'Capt': 'Rareman', 'Col': 'Rareman', 'Don': 'Rareman', 'Dona': 'Rarewoman',
      'Dr': 'Rareman', 'Jonkheer': 'Rareman', 'Lady': 'Rarewoman', 'Major': 'Rareman',
      'Master': 'Master', 'Miss': 'Miss', 'Mlle': 'Rarewoman', 'Mme': 'Rarewoman',
      'Mr': 'Mr', 'Mrs': 'Mrs', 'Ms': 'Rarewoman', 'Rev': 'Mr', 'Sir': 'Rareman',
      'the Countess': 'Rarewoman'}
full.Title = full.Title.map(nn)
# assign the female 'Dr' to 'Rarewoman'
full.loc[full.PassengerId == 797, 'Title'] = 'Rarewoman'
full.Age.fillna(999, inplace=True)


def girl(aa):
    if (aa.Age != 999) & (aa.Title == 'Miss') & (aa.Age <= 14):
        return 'Girl'
    elif (aa.Age == 999) & (aa.Title == 'Miss') & (aa.Parch != 0):
        return 'Girl'
    else:
        return aa.Title


full['Title'] = full.apply(girl, axis=1)

Tit = ['Mr', 'Miss', 'Mrs', 'Master', 'Girl', 'Rareman', 'Rarewoman']
for i in Tit:
    full.loc[(full.Age == 999) & (full.Title == i), 'Age'] = full.loc[full.Title == i, 'Age'].median()

full.loc[full['Age'] <= 15, 'Age'] = 0
full.loc[(full['Age'] > 15) & (full['Age'] < 55), 'Age'] = 1
full.loc[full['Age'] >= 55, 'Age'] = 2
full['Pclass'] = full['Pclass'] - 1


Fare = full['Fare'].values
Fare = Fare.reshape(-1, 1)
km = KMeans(n_clusters=3).fit(Fare)  # 将数据集分为2类
Fare = km.fit_predict(Fare)
full['Fare'] = Fare
full['Fare'] = full['Fare'].astype(int)
full['Age'] = full['Age'].astype(int)
full['Cabin'] = full['Cabin'].astype(int)
full['Pclass'] = full['Pclass'].astype(int)
full['Sex'] = full['Sex'].astype(int)
# full['Survived']=full['Survived'].astype(int)
print(full)    # 1308行

dataset = full.drop(columns=['Embarked', 'Name', 'Parch', 'PassengerId', 'SibSp', 'Ticket', 'Title'])
print(dataset)    # 1308行
test_data = dataset[891:]
dataset.dropna(inplace=True)
print(dataset)    # 891行
dataset['Survived'] = dataset['Survived'].astype(int)
# dataset=pd.concat([dataset, pd.DataFrame(columns=['Pri'])])
train = dataset[:800]
val = dataset[800:]
'''
最后保留如下项目,并切出800的训练集：
Pclass => 客舱等级(0/1/2等舱位)
Sex => 性别 male=1 female=0
Age => 年龄 插补后分0,1,2 代表 幼年（0-15） 成年（15-55） 老年（55-）
Fare => 船票价格 经聚类变0 1 2 代表少 多 很多
Cabin => 客舱号 清洗成有无此项，并发现有的生存率高
'''

#model = BayesianModel([('Age', 'Pri'), ('Sex', 'Pri'),('Pri','Survived'),('Fare','Pclass'),('Pclass','Survived'),('Cabin','Survived')])
model = BayesianModel([('Age', 'Survived'), ('Sex', 'Survived'),('Fare','Pclass'),('Pclass','Survived'),('Cabin','Survived')])

model.fit(train, estimator=BayesianEstimator, prior_type="BDeu") # default equivalent_sample_size=5

# #输出节点信息
# print(model.nodes())
# #输出依赖关系
# print(model.edges())
# #查看某节点概率分布
# print(model.get_cpds('Survived').values)

val_data=val.drop(columns=['Survived'], axis=1)
print(val_data)
y_pred = model.predict(val_data)
val = val.reset_index(drop=True)    # 解决以下行索引不同问题
print((y_pred['Survived']==val['Survived']).sum()/len(val))
#测试集精度0.8131868131868132

# 预测test.csv中人员的存活情况
predict_data = test_data.drop(columns=['Survived'], axis=1)
predict_pred = model.predict(predict_data)
predict_pred.to_csv('gender_submission.csv')

这里将竞赛数据集中的训练集分成训练集(train)和验证集(val)，竞赛数据集中的测试集作为测试集。输出结果：

将预测的gender_submission.csv提交到kaggle中，获得了如下所示准确率。

参考

[1] 概率图模型理论与应用–哔哩哔哩
[2] 贝叶斯网络python实战（以泰坦尼克号数据集为例，pgmpy库）–CSDN