CODEWITHSUNDEEP
pythonmachine-learningscikit-learndecision-tree
blue-sky
blue-sky

Reputation: 53786

Visualizing decision tree : IndexError: list index out of range

Below decision tree :

enter image description here

Is generated using code : 

dt = DecisionTreeClassifier()
dt = clf.fit([[1],[2],[3]], [[3],[2],[3]])

dot_data = export_graphviz(dt, out_file=None, 
                         feature_names=['1' , '2' , '3'],  
                         class_names=['true' , 'false'],  
                         filled=True, rounded=True,  
                         special_characters=True)  
graph = graphviz.Source(dot_data)  
graph

If I use instead : 

dt = DecisionTreeClassifier()
dt = clf.fit([[1],[2],[3]], [[2],[3],[4]])

dot_data = export_graphviz(dt, out_file=None, 
                         feature_names=['1' , '2', '3'],  
                         class_names=['true' , 'false'],  
                         filled=True, rounded=True,  
                         special_characters=True)  
graph = graphviz.Source(dot_data)  
graph

error is returned : 

/usr/local/lib/python3.5/dist-packages/sklearn/tree/export.py in node_to_str(tree, node_id, criterion)
    284                 node_string += 'class = '
    285             if class_names is not True:
--> 286                 class_name = class_names[np.argmax(value)]
    287             else:
    288                 class_name = "y%s%s%s" % (characters[1],

IndexError: list index out of range

Is this a quirk of the visualization as the classifier trains correctly ?

Upvotes: 0

Views: 10375

Answers (1)

Gambit1614
Gambit1614

Reputation: 8801

I think there is one typo in your code. You are using clf.fit whereas it should be dt.fit.

Secondly, in the second instance you are specifying three labels, namely ['2','3','4'] whereas you are specifying the class labels as only true and false which is wrong, since you are clearly using more than 2 labels and it can't map an array of size 2 (class names) to an array of size 3(your actual labels). So basically you need to add another label besides true and false and it should work correctly.

dt = DecisionTreeClassifier()
dt = dt.fit([[1],[2],[3]], [[2],[3],[4]])   #It should be dt.fit not clf.fit

dot_data = export_graphviz(dt, out_file=None, 
                     feature_names=['1' , '2', '3','4'],  
                     class_names=['true' , 'false','something_else'],  
                     filled=True, rounded=True,  
                     special_characters=True)

Now it should work correctly. Feel free to name the third label as you like. The error basically occurred since you didn't specify anything for the third label, hence it could not map the actual labels (i.e. 2,3 and 4) to the ones you specified in the class_names, i.e. true and false.

Upvotes: 3

Related Questions