access to numbers in classification_report - sklearn

Question

This is a simple example of a classification_report in sklearn:

from sklearn.metrics import classification_report
y_true = [0, 1, 2, 2, 2]
y_pred = [0, 0, 2, 2, 1]
target_names = ['class 0', 'class 1', 'class 2']
print(classification_report(y_true, y_pred, target_names=target_names))
#             precision    recall  f1-score   support
#
#    class 0       0.50      1.00      0.67         1
#    class 1       0.00      0.00      0.00         1
#    class 2       1.00      0.67      0.80         3
#
#avg / total       0.70      0.60      0.61         5

I want to have access to avg/total row. For instance, I want to extract the f1-score from the report, which is 0.61.

How can I have access to the number in classification_report?

Joy

are you interested in the f1-score or extracting f1-score from classification report? – Lannie

@PratikKumar extracting from classification report. I need other reports also. – Joy

Answer 1

You can output the classification report by adding output_dict=True to the report:

report = classification_report(y_true, y_pred, output_dict=True)

And then access its single values as in a normal python dictionary.

For example, the macro metrics:

macro_precision =  report['macro avg']['precision'] 
macro_recall = report['macro avg']['recall']    
macro_f1 = report['macro avg']['f1-score']

or Accuracy:

accuracy = report['accuracy']

Nonmetallic

Answer 2

you can use precision_recall_fscore_support for getting all at once

from sklearn.metrics import precision_recall_fscore_support as score
y_true = [0, 1, 2, 2, 2]
y_pred = [0, 0, 2, 2, 1]
precision,recall,fscore,support=score(y_true,y_pred,average='macro')
print 'Precision : {}'.format(precision)
print 'Recall    : {}'.format(recall)
print 'F-score   : {}'.format(fscore)
print 'Support   : {}'.format(support)

here is the link to the module

Lannie

The answer is correct, but please note that you have used the wrong parameters, since the first parameter is y_true, the second one should be y_pred. – Creath

Does this also work for multi-class datasets? – Holloweyed

Answer 3

You can use output_dict parameter in build-in classification_report to return a dictionary:

classification_report(y_true,y_pred,output_dict=True)

Impinge

Answer 4

classification_report is string so I would suggest you to use f1_score from scikit-learn

from sklearn.metrics import f1_score
y_true = [0, 1, 2, 2, 2]
y_pred = [0, 0, 2, 2, 1]
target_names = ['class 0', 'class 1', 'class 2']

print(f1_score(y_true, y_pred, average=None)

output

Walliw

Thank you. so there is no way to extract from classification_report? what about the other reports? – Joy

maybe you can use regex to extract this value. can you name the other reports ? – Walliw

If you are talking about recall and precision, yes there are functions like recall_score and precision_score in sklearn – Walliw

Answer 5

You have to set the parameter output_dict to True it's not mentioned in the documentation but it is in repo L2636

Example

report = classification_report(y_true, y_pred, 
                               output_dict=True) #<== Right here

This will output the dict with keys report.keys() the output is;

dict_keys(['0', '1', '2', 'accuracy', 'macro avg', 'weighted avg'])

Dowel