【深耕 Python】Data Science with Python 数据科学（17）Scikit-learn机器学习（二）

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

URL = "https://learnenough.s3.amazonaws.com/titanic.csv"
titanic = pd.read_csv(URL)

二、导入机器学习模型

Scikit-learn提供的机器学习模型（部分，附介绍链接）：

逻辑斯蒂回归 Logistic Regression

朴素贝叶斯 Naive Bayes

感知机 Perceptron

决策树 Decision Tree

随机森林 Random Forest

导入上述机器学习模型：

python 复制代码

from sklearn.linear_model import LogisticRegression
from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import Perceptron
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier

三、数据预处理

将除了舱级Pclass 、性别Sex 、年龄Age 和生还Survived之外的列全部去除：

python 复制代码

dropped_columns = ["PassengerId", "Name", "Cabin", "Embarked", "SibSp", "Parch", "Ticket", "Fare"]
for column in dropped_columns:
    titanic = titanic.drop(column, axis=1)

然后，将剩余列中的NaN 和NaT值去除：

python 复制代码

for column in ["Age", "Sex", "Pclass"]:
    titanic = titanic[titanic[column].notna()]

还需将分类变量（Categorical Variable），比如性别，映射为数值变量（Numerical Variable）：

python 复制代码

sexes = {"male": 0, "female": 1}
titanic["Sex"] = titanic["Sex"].map(sexes)

准备自变量和因变量：

python 复制代码

X = titanic.drop("Survived", axis=1)
Y = titanic["Survived"]

观察自变量和因变量的数据结构：

python 复制代码

print(X.head(), "\n----\n")
print(Y.head(), "\n----\n")

程序输出：

python 复制代码

# 3个自变量
   Pclass  Sex   Age
0       3    0  22.0
1       1    1  38.0
2       3    1  26.0
3       1    1  35.0
4       3    0  35.0 
----

# 因变量
0    0
1    1
2    1
3    1
4    0
Name: Survived, dtype: int64 
----

接下来，将原数据划分为训练集和测试集，需导入 train_test_split() 方法：

python 复制代码

from sklearn.model_selection import train_test_split

(X_train, X_test, Y_train, Y_test) = train_test_split(X, Y, random_state=1)

四、定义、训练和评估模型

逻辑斯蒂回归模型

python 复制代码

logreg = LogisticRegression()
logreg.fit(X_train, Y_train)
accuracy_logreg = logreg.score(X_test, Y_test)

（高斯）朴素贝叶斯模型

python 复制代码

naive_bayes = GaussianNB()
naive_bayes.fit(X_train, Y_train)
accuracy_naive_bayes = naive_bayes.score(X_test, Y_test)

感知机模型

python 复制代码

perceptron = Perceptron()
perceptron.fit(X_train, Y_train)
accuracy_perceptron = perceptron.score(X_test, Y_test)

决策树模型

python 复制代码

decision_tree = DecisionTreeClassifier()
decision_tree.fit(X_train, Y_train)
accuracy_decision_tree = decision_tree.score(X_test, Y_test)

随机森林模型

python 复制代码

random_forest = RandomForestClassifier()
random_forest.fit(X_train, Y_train)
accuracy_random_forest = random_forest.score(X_test, Y_test)

模型评估：

python 复制代码

results = pd.DataFrame({
    "Model": ["Logistic Regression", "Naive Bayes", "Perceptron",
              "Decision Tree", "Random Forest"],
    "Score": [accuracy_logreg, accuracy_naive_bayes, accuracy_perceptron,
              accuracy_decision_tree, accuracy_random_forest]
})

result_df = results.sort_values(by="Score", ascending=False)
result_df = result_df.set_index("Score")
print(result_df)

模型准确率：

python 复制代码

# 准确率        模型               
Score           Model                        
0.854749        Decision Tree
0.854749        Random Forest
0.787709  Logistic Regression
0.770950          Naive Bayes
0.743017           Perceptron

对随机森林模型中的3个因素的权重进行分析并绘制柱状图：

python 复制代码

print(random_forest.feature_importances_)
print(X_train.columns)
fig, ax = plt.subplots()
ax.bar(X_train.columns, random_forest.feature_importances_)
plt.title("Factor Importance of Random Forest")
plt.ylabel("Importance")
plt.grid()
plt.show()

程序输出：

python 复制代码

[0.17858357 0.35377705 0.46763938]
Index(['Pclass', 'Sex', 'Age'], dtype='object')

五、交叉验证

对随机森林模型进行K折交叉验证（默认值为K=5):

python 复制代码

from sklearn.model_selection import cross_val_score

random_forest = RandomForestClassifier(random_state=1)
scores = cross_val_score(random_forest, X, Y)
print(scores)
print(scores.mean())
print(scores.std())

程序输出：

python 复制代码

[0.75524476 0.8041958  0.82517483 0.83216783 0.83098592]  # 5次交叉验证
0.8095538264552349  # 平均准确率
0.028958338744358988  # 标准差

参考文献 Reference

《Learn Enough Python to be Dangerous------Software Development, Flask Web Apps, and Beginning Data Science with Python》, Michael Hartl, Boston, Pearson, 2023.