VsCode舒适系列之graphviz决策树可视化

接到小伙伴的决策树代码实现需求,为了在windows环境下愉快的调试并输出结果，调整了下环境细节。现在将整个过程记录并分享

环境

• win10
• VsCode
• GitHub Desktop
• python3.8

  VsCode配置

1. vscode配置ptyhon
   1. Ctrl + Shift + X 打开应用商店搜索安装python就OK了
2. vscode配置jupyter
3. vscode配置graphviz
   • 下载graphviz
   • 配置graphviz的/bin目录到windows环境变量
   • 然后执行pip install graphviz
   • 最后在配置好的jupyter中执行代码就能享受可视化了

     参考

GitHub Desktop配置

vscode的github插件有点感人，安利个GitHub Desktop

• 下载、登录~~似乎没啥介绍的
• 便捷同步github库、打开IDE、排除同步文件、快速建代码仓
• Ctrl+ P同步、Ctrl+ Shift + A …. 必要快捷键，界面上有提示
• 

  通过jupyter导出的代码

  tips: jupyter调试好的code可以直接通过角标导出

# To add a new cell, type '# %%'
# To add a new markdown cell, type '# %% [markdown]'
# %%
import numpy as np
import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
from sklearn import tree
from sklearn.tree import DecisionTreeClassifier as DTC, export_graphviz
import pydot
import pydotplus
import time
from IPython.display import Image
# %%
Train_data = pd.read_csv('./datalab/car_info_test.csv')
Train_data.info()
# %%
Train_data
# %%
# 筛选数值特征列
numerical_cols = Train_data.select_dtypes(exclude = 'object').columns
print(numerical_cols)
# %%
# 对部分特征列进行分类
Train_data['CUST_AGE'] = Train_data.CUST_AGE.apply(lambda x: 1 if (x<=35 and x >=16) else 2 if (x >= 36 and x <= 60) else 3)
Train_data['CAR_AGE'] = Train_data.CAR_AGE.apply(lambda x: 1 if (x <=730) else 2 if (x >=731 and x <= 1460) else 3 if (x >= 1461 and x <= 2190) else 4 if (x >= 2191 and x <= 3650) else 5)
Train_data['CAR_PRICE'] = Train_data.CAR_PRICE.apply(lambda x: 1 if (x >=50000 and x <= 90000) else 2 if (x >= 90001 and x<= 150000) else 3 if (x >= 150001 and x <= 300000) else 4)
Train_data['LOAN_AMOUNT'] = Train_data.LOAN_AMOUNT.apply(lambda x: 1 if (x <= 50000) else 2 if(x >=50001 and x <= 200000) else 3 if(x >= 200001 and x <= 500000) else 4)
# %%
Train_data.head()
# %%
# 查看是否有空值特征列
Train_data.isnull().any()
# %%
# 从数值特种列中区分 data 和 taeget（标记的特征）
feature_cols = [col for col in numerical_cols if col != 'IS_LOST']
data = Train_data[feature_cols].fillna(value=0)
target = Train_data['IS_LOST']
# %%
# DecisionTreeClassifier 的决策树实现
dtc = DTC(criterion='entropy',max_depth=6)
dtc.fit(data, target)
print('准确率:', dtc.score(data, target))
# %%
# 可视化展示
dot_data = tree.export_graphviz(dtc, out_file=None, 
                         filled=True, rounded=True,  
                         special_characters=True)  
graph = pydotplus.graph_from_dot_data(dot_data)  
Image(graph.create_png()) 
# %%
# 输出为PDF
import pydot
import pydotplus
with open('./tree.dot', 'w') as f:
    f = export_graphviz(dtc, feature_names=data.columns, out_file=f)
graph = pydotplus.graph_from_dot_file('./tree.dot')
graph.write_pdf("iris.pdf")

效果展示

听说vim能舒爽到刷剧