Python决策树入门案例:
决策树是机器学习中一个比较重要而且常用的算法, 是基于香农的信息论计算信息熵然后计算信息增益
然后划分决策树的"枝叶
Python实现计算信息熵公式:
from fractions import Fraction # 导入分数计算模块
from math import log
a = Fraction(4, 6) # 正例占3/6
b = Fraction(2, 6) # 反例占2/6
Ent(D) = -(a * log(a, 2) + b * log(b, 2)) # |Y| = 2
import numpy as np
import pandas as pd
file_path = "E:\\数据集\\练习数据集\\titanic.csv"
data = pd.read_csv(file_path)
data["age"].isnull().sum() # 年龄有680个缺失值
x = data[["pclass", "age", "sex"]]
y = data["survived"]
x["age"].fillna(x["age"].mean(), inplace=True) # 用平均年龄来填充缺失值
from sklearn.model_selection import train_test_split # 导入数据集分割
from sklearn.feature_extraction import DictVectorizer # 导入特征工程
from sklearn.tree import DecisionTreeClassifier # 导入决策树分类器
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.25) # 分割数据
info = DictVectorizer(sparse=False) # 特征工程
x_train = info.fit_transform(x_train.to_dict(orient="records"))
x_test = info.fit_transform(x_test.to_dict(orient="records"))
dec = DecisionTreeClassifier(max_depth=5)
dec.fit(x_train, y_train)
dec.score(x_test, y_test)
dec.predict(x_test[0: 1]) # 进行测试预测
import pydotplus # 可视化
dot_data = tree.export_graphviz(dec, out_file=None,
filled=True, rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data)
graph.get_nodes()[7].set_fillcolor("#FFF2DD")
graph.write_png("graph.png")
from IPython.display import Image
Image(graph.create_png())
构造的决策树如下:
数据集地址: http://biostat.mc.vanderbilt.edu/wiki/pub/Main/DataSets/titanic.txt