- Visit this site for the original codes.
- Start your jupyter notebook ジュピターノートブックを開始する
- (1) Start your "Anaconda navigator" then select jupyter notebook. アナコンダ・ナビゲータを立ち上げて、そのメニューからジュピターノートブックを立ち上げてもよい
- (2) Or start "jupyter notebook" directly. You can type "jupyter notebook" in your terminal for this. ジュピターノートブックを直接立ち上げてもよい。ターミナルに"jupyter notebook"と打ち込んで立ち上げることもできる
- Open a new jupyter notebook with "python3" kernel in your browser. 新しいジュピターノートブック(カーネルはpython3)をブラウザで開く
- Write (or copy-and-paste) the following python commands 以下のパイソンコマンドをノートブックに書きます(コピーペーストします)
- Each section should be filled in each chunk. 一区切りごとに１チャンクに書き込みます
- We will encounter various errors, but let's solve them one by one. エラーが続出すると思いますが、一つずつ解決していきます
import pandas as pd
import numpy as np
from sklearn.tree import DecisionTreeClassifier, export_graphviz
import graphviz
import pydotplus
from IPython.display import Image
from sklearn.externals.six import StringIO
data = pd.DataFrame({
"buy(y)":[True,True,True,True,True,True,True,False,False,False,False,False,False],
"high":[4, 5, 3, 1, 6, 3, 4, 1, 2, 1, 1,1,3],
"size":[30, 45, 32, 20, 35, 40, 38, 20, 18, 20, 22,24,25],
"autolock":[1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0,1,0]
})
y = data.loc[:,["buy(y)"]]
X = data.loc[:,["high", "size","autolock"]]
clf = DecisionTreeClassifier()
clf = clf.fit(X, y)
dot_data = StringIO()
export_graphviz(clf, out_file=dot_data,
feature_names=["high", "size","autolock"],
class_names=["False","True"],
filled=True, rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png())
dot_data = StringIO()
export_graphviz(clf, out_file=dot_data,
feature_names=["high", "size","autolock"],
class_names=["False","True"],
filled=True, rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
graph.progs = {'dot': u"C:\\tools\\Anaconda3\\Library\\bin\\dot.bat"}
Image(graph.create_png())
print(graph.to_string())
- pip
- Open the regular terminal
pip install --upgrade pip
-
- Install the packages if you get errors @ import xxxx
pip install xxxx
graphviz.org
-
- Graphviz can be installed in your local PC, either Mac or Windows... ダウンロードしてMac/WindowsのローカルPC上でも使えます
- In my environment, python3 on jupyter notebook was able to call the installed Graphviz without further settings.
- In my environment, python3 on jupyter notebook failed to call the installed Graphviz (probably "path-problem"). Therefore, a python code line was added to specify the location of installed Graphviz's "dot"-functionality command file "dot.bat", based on an instruction here
niwakomablog.com.
conda install graphviz
-
-
- It installed "Graphviz" under the Anaconda folder in my windows PC: "C:\\tools\\Anaconda3\\Library\\bin\\dot.bat" WindowsではそのコマンドによりAnaconda関連の一式を納めた領域にGraphvizがインストールされます。
- As far as searched the websites, various errors/troubles around Graphviz and python connection. Graphvizをpythonで使うためにはいろいろな方法でGraphvizをインストールすることができるようです(難航した人が結構、存在する、ということの裏返し)
- If no solution, print out "dot-formatted string" and copy-and-paste the contents to the Graphviz-online. もしpythonからのGraphviz呼び出しに成功しなかったり、そもそもGraphvizがローカルPCで動かない場合には、pythonでGraphvizが受け付ける「グラフを描くための文字列」を作って、それを「Graphvizをウェブサイト上で動かすサイト」に持っていくという手もあります
dreampuf.github.io
print(graph.to_string())
digraph Tree {
node [color="black", fontname=helvetica, shape=box, style="filled, rounded"];
edge [fontname=helvetica];
0 [fillcolor="#399de524", label=<size ≤ 27.5<br/>gini = 0.497<br/>samples = 13<br/>value = [6, 7]<br/>class = True>];
1 [fillcolor="#e58139d4", label=<autolock ≤ 0.5<br/>gini = 0.245<br/>samples = 7<br/>value = [6, 1]<br/>class = False>];
0 -> 1 [headlabel="True", labelangle=45, labeldistance="2.5"];
2 [fillcolor="#e58139ff", label=<gini = 0.0<br/>samples = 5<br/>value = [5, 0]<br/>class = False>];
1 -> 2;
3 [fillcolor="#e5813900", label=<size ≤ 22.0<br/>gini = 0.5<br/>samples = 2<br/>value = [1, 1]<br/>class = False>];
1 -> 3;
4 [fillcolor="#399de5ff", label=<gini = 0.0<br/>samples = 1<br/>value = [0, 1]<br/>class = True>];
3 -> 4;
5 [fillcolor="#e58139ff", label=<gini = 0.0<br/>samples = 1<br/>value = [1, 0]<br/>class = False>];
3 -> 5;
6 [fillcolor="#399de5ff", label=<gini = 0.0<br/>samples = 6<br/>value = [0, 6]<br/>class = True>];
0 -> 6 [headlabel="False", labelangle="-45", labeldistance="2.5"];
}
- Regular terminal & Anaconda terminal
- You can get the jupyter notebook file here: 以上のpythonコマンドをjupyter notebook形式で保存したものは、以下から取れます
github.com
