解决sklearn，Graphviz画决策树文乱码的问题

1.  
   from sklearn import tree
2.  
   from sklearn.datasets import load_wine
3.  
   from sklearn.model_selection import train_test_split
4.  
   import pydotplus
5.  
   from IPython.display import Image
6.  
    
7.  
   wine = load_wine()
8.  
   Xtrain,Xtest,Ytrain,Ytest = train_test_split(wine.data,wine.target,test_size=0.3)
9.  
   clf = tree.DecisionTreeClassifier(criterion='entropy')
10.  
    clf = clf.fit(Xtrain,Ytrain)
11.  
    score = clf.score(Xtest,Ytest)
12.  
    print(score)
13.  
     
14.  
     
15.  
    feature_name = ['酒精','苹果酸','灰','灰的碱性','镁','总酚','类黄酮','非黄烷类酚类','花青素','颜色强度','琴酒','雪莉','贝尔摩德']
16.  
    dot_data = tree.export_graphviz(clf
17.  
    ,out_file=None
18.  
    ,feature_names= feature_name
19.  
    ,class_names=['琴酒','雪莉','贝尔摩德']
20.  
    ,filled = True
21.  
    ,rounded = True
22.  
    ,special_characters=True)
23.  
     
24.  
    graph = pydotplus.graph_from_dot_data(dot_data)
25.  
    Image(graph.create_png())
26.  
    graph.write_png("wine0.png")

将F:\python387\Lib\site-packages\sklearn\tree\_export.py中的“helvetica”改为“simsun”或‘Micorsoft YaHei’ 即可。在程序运行时需要在 filled = True 和 special_characters=True 之间加上rounded = True，即可。

这篇好文章是转载于：学新通技术网