sklearn实现决策树并可视化+linux系统

失败的历程

pydot+graphviz

tree.export_graphviz(regr_1, out_file=f)
(graph,) = pydot.graph_from_dot_file("./tree.dot")
graph.write_png("./tree.png")

然后就是一直报错找不到dot文件，一直报，一直报，报到人头铁了都。
我甚至取读了pydot.py的源码，报错信息应该来源于一下几行代码：

我甚至根据网上的建议改了下面的函数

然后就什么卵用都没有~~~~

成功的做法

pydotplus+graphviz

过程如下：

第一步：安装相应的工具

pip install pydotplus
pip install graphviz

上面的一般不会出现什么问题，关键是下面这一个graphviz
在自己的机器上也还要有个这玩意，linux安装就比较方便：

yum install graphviz

第二步：整体过程

import pandas as pd
import numpy as np
from sklearn import tree
import graphviz
import pydot
import pydotplus
from sklearn.metrics import mean_squared_error,mean_absolute_error
from sklearn.tree import DecisionTreeRegressor
X_train=pd.read_csv("../state/data/trainData/X_train.csv",index_col=0)
X_test=pd.read_csv("../state/data/trainData/X_test.csv",index_col=0)
Y_train=pd.read_csv("../state/data/trainData/Y_train.csv",index_col=0)
Y_test=pd.read_csv("../state/data/trainData/Y_test.csv",index_col=0)

X_train_array=np.asarray(X_train)
X_test_array=np.asarray(X_test)
Y_train_array=np.asarray(Y_train)
Y_test_array=np.asarray(Y_test)
Y_train_array.reshape(len(Y_train_array),-1)
Y_test_array.reshape(len(Y_test_array),-1)
for dep in [3]:
    regr_1 = DecisionTreeRegressor(max_depth=dep)
    # criterion = "mse",参数选取的规则，回归的时候可以选mse或者mae，对于分类问题选择Gini系数或者Entropy分别对应CART算法和C4.5s
    # splitter = "best",是随机的选取划分点还是随机的选择划分点。
    # max_depth = None,树的最大的深度
    # min_samples_split = 2,在分割过程中考虑的最少的样本，默认是2
    # min_samples_leaf = 1,叶子节点的最少样本
    # min_weight_fraction_leaf = 0.,
    # max_features = None,
    # random_state = None,
    # max_leaf_nodes = None,
    # min_impurity_decrease = 0.,
    # min_impurity_split = None,
    # presort = 'deprecated',
    # ccp_alpha
    regr_1=regr_1.fit(X_train_array, Y_train_array)
    y_pre=regr_1.predict(X_test_array)

    mse=mean_squared_error(y_pre,Y_test_array)
    mae=mean_absolute_error(y_pre,Y_test_array)
    print("树的深度",dep," mae: ",mae," mse ",mse)
    with open("tree.dot", 'w') as f:
        f = tree.export_graphviz(regr_1, out_file=f)
    # (graph,) = pydot.graph_from_dot_file("./tree.dot")
    # (graph,) = pydotplus.graph_from_dot_file("./tree.dot")
    dot_data = tree.export_graphviz(regr_1, out_file=None,
                                    # feature_names=iris.feature_names,
                                    # class_names=iris.target_names,
                                    filled=True, rounded=True,
                                    special_characters=True)
    graph=pydotplus.graph_from_dot_data(dot_data)
    graph.write_png("./tree.png")

    #
    # graph.save()
    # dot.view()

舒服了~