决策树DecisionTree

模型亮点

1. 初始测试集上评分为0.4，调参后测试集上评分为0.8
2. 数据清洗方式得当

-----------------------------------------以下为模型具体实现-----------------------------------------

Step1.数据读取

import pandas as pd
df=pd.read_csv('bankdebt.csv',index_col=0,header=None)
df.columns=['house','marital','income','repayment']
df.head()

Step2.数据清洗

why 数据清洗？how 数据清洗？

1. 决策树不要求标准化
2. 分类数据，用0-1表示
3. 顺序数据，用1、2、3表示

# 1.是否有房子，Yes->1，No->0
df.loc[df['house']=='Yes','house']=1
df.loc[df['house']=='No','house']=0

# 2.能否偿还贷款，Yes->1，No->0
df.loc[df['repayment']=='Yes','repayment']=1
df.loc[df['repayment']=='No','repayment']=0

# 3.婚姻状态，Single->1，Married->2，Divorced->3
df.loc[df['marital']=='Single','marital']=1
df.loc[df['marital']=='Married','marital']=2
df.loc[df['marital']=='Divorced','marital']=3

Step3.划分训练集和测试集

why 划分训练集和测试集？

• 把所有样本当作训练集，做过的题都是旧题，都会~
• 把部分样本当作训练集，在新题上做测试，起到检测效果~

from sklearn.model_selection import train_test_split
x=df.drop('repayment',axis=1).astype(float)
y=df['repayment'].astype(float)
x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.3,random_state=1)

Step4.启动决策树

how 启动决策树？

1. 设定初始化参数
2. 拟合训练集数据

from sklearn.tree import DecisionTreeClassifier
def tree(n,x_train,y_train):
    model=DecisionTreeClassifier(max_depth=n)
    model.fit(x_train,y_train)
    return model
model=tree(1,x_train,y_train) #初始决策树，max_depth设为1

Step5.模型评估

what 模型评估？

• 分类问题，model.score->accuracy_score
• 回归问题，model.score->r2_score

print("-----初始决策树-----")
print("训练集上评分：",model.score(x_train,y_train))
def score(model,x_test,y_test):
    print("测试集上评分：",model.score(x_test,y_test))
score(model,x_test,y_test)

Step6.优化参数

why 优化参数？how 优化参数？

1. 以提高测试集上评分
2. 参数max_depth，对分类效果影响较大

for n in range(2,11):
    print("-----决策树最大深度为",n,"-----")
    model=tree(n,x_train,y_train)
    score(model,x_test,y_test)

Step7.可视化决策树结构

how 可视化决策树结构？

• https://blog.csdn.net/m0_37634594/article/details/122309924

from sklearn.tree.export import export_text
branch_name=['house','marital','income']
export_text(model,feature_names=branch_name)

Step8.保存模型

why 保存模型？how 保存模型？

1. for job-lib（工作自由）~
2. dump转存模型，以pkl格式
3. load加载模型

from sklearn.externals import joblib
joblib.dump(model,'d:\DecisionTree.pkl')
new_model=joblib.load('d:\DecisionTree.pkl')
print("测试集上预测结果为：\n",new_model.predict(x_test))

-END