Question

我想用export_graphviz可视化我的决策树，但是我不断遇到以下错误：


  File "C:\Users\User\AppData\Local\Continuum\anaconda3\envs\data_science\lib\site-packages\sklearn\utils\validation.py", line 951, in check_is_fitted
    raise NotFittedError(msg % {'name': type(estimator).__name__})

NotFittedError: This Pipeline instance is not fitted yet. Call 'fit' with appropriate arguments before using this method.

我非常确定我的管道是否合适，因为我在代码中调用了predict，效果很好。这是有问题的代码：

from sklearn.tree import DecisionTreeRegressor
import pandas as pd
import numpy as np
from sklearn.pipeline import Pipeline
from sklearn.model_selection import train_test_split
from sklearn.tree import export_graphviz

#Parameters for model building an reproducibility
state = 13

data_age.dropna(inplace=True)
X_age = data_age.iloc[:,0:77]
y_age =  data_age.iloc[:,77]

X = X_age
y = y_age

#split between testing and training set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state= state)

# Pipeline with the regressor
regressors = [DecisionTreeRegressor(random_state = state)]
for reg in regressors:
    steps=[('regressor', reg)]
    pipeline = Pipeline(steps) #seed that controls the random grid search

#Train the model

pipeline.set_params(regressor__max_depth = 5, regressor__min_samples_split =5, regressor__min_samples_leaf = 5).fit(X_train, y_train)
pred = pipeline.predict(X_test)
pipeline.score(X_test, y_test)

export_graphviz(pipeline, out_file='tree.dot')

我知道我真的不需要这里的管道，但是我仍然想了解问题所在以备将来参考，并能够在已安装管道的情况下绘制决策树。

Answer 1

REP的签名为export_graphviz，如documentation所示。

因此，您应该将决策树作为参数传递给export_graphviz(decision_tree, ...)函数，而不是管道。

您还可以在source code中看到export_graphviz正在调用export_grpahviz方法。

Answer 2

因此，根据Farseer的回答，最后一行必须为：


#Train the model
pipeline.set_params(regressor__max_depth = 5, regressor__min_samples_split =5, regressor__min_samples_leaf = 5).fit(X_train, y_train)
pred = pipeline.predict(X_test)
pipeline.score(X_test, y_test)

#export as a .dot file
export_graphviz(regressors[0], out_file='tree.dot')

现在可以使用了。

管道内决策树的可视化

2 个答案: