如何使用 python 打印精度、召回率、fscore？

Question

I want to calculate and print precision, recall, fscore and support using sklearn.metrics in python. I am doig NLP so my y_test and y_pred are basicaly words before the vectorisation step.

below some information that can help you :

y_test:  [0 0 0 1 1 0 1 1 1 0]
y_pred [0.86 0.14 1.   0.   1.   0.   0.04 0.96 0.01 0.99 1.   0.   0.01 0.99
 0.41 0.59 0.02 0.98 1.   0.  ]

x_train 50
y_train 50
x_test 10
y_test 10
x_valid 6
y_valid 6

y_pred dimension:  (20,)
y_test dimension:  (10,)

the full trackback error :

  Traceback (most recent call last):
  File "C:\Users\iduboc\Documents\asd-dev\train.py", line 324, in <module>
    precision, recall, fscore, support = score(y_test, y_pred)
  File "C:\Users\iduboc\Python1\envs\asd-v3-1\lib\site-packages\sklearn\metrics\classification.py", line 1415, in precision_recall_fscore_support
    pos_label)
  File "C:\Users\iduboc\Python1\envs\asd-v3-1\lib\site-packages\sklearn\metrics\classification.py", line 1239, in _check_set_wise_labels
    y_type, y_true, y_pred = _check_targets(y_true, y_pred)
  File "C:\Users\iduboc\Python1\envs\asd-v3-1\lib\site-packages\sklearn\metrics\classification.py", line 71, in _check_targets
    check_consistent_length(y_true, y_pred)
  File "C:\Users\iduboc\Python1\envs\asd-v3-1\lib\site-packages\sklearn\utils\validation.py", line 205, in check_consistent_length
    " samples: %r" % [int(l) for l in lengths])
ValueError: Found input variables with inconsistent numbers of samples: [10, 20]

my code :

 from sklearn.metrics import precision_recall_fscore_support as score
    precision, recall, fscore, support = score(y_test, y_pred)
    print('precision: {}'.format(precision))
    print('recall: {}'.format(recall))
    print('fscore: {}'.format(fscore))
    print('support: {}'.format(support))

My code to predict the values :

elif clf == 'rndforest':

    # No validation data in rnd forest
    x_train = np.concatenate((x_train, x_valid))
    y_train = np.concatenate((y_train, y_valid))

    model = RandomForestClassifier(n_estimators=int(clf_params['n_estimators']),
                                   max_features=clf_params['max_features'])
    model.fit(pipe_vect.transform(x_train), y_train)

    datetoday = datetime.today().strftime('%d-%b-%Y-%H_%M')
    model_name_save = abspath(os.path.join("models", dataset,  name_file + '-' + 
    vect + reduction + '-rndforest'\
                                   + datetoday + '.pickle'))
    print("Model d'enregistrement : ", model_name_save)




    x_test_vect = pipe_vect.transform(x_test)

    y_pred = model.predict_proba(x_test_vect)  

Answer 1

The error is due to the different sizes of the predicted and ground truth vectors. The function precision_recall_fscore_support only works if these sizes are the same.

See the docs:

https://scikit-learn.org/stable/modules/generated/sklearn.metrics.precision_recall_fscore_support.html

Also, the aforementioned function expects to receive non-continuous values, otherwise. If you pass as an argument a list with floats between 0 and 1 ( y_pred list) you will have the next error:

ValueError: Classification metrics can't handle a mix of binary and continuous targets

The example code that produced the error is this:

y_test =  [0., 0., 0., 1., 1.]
y_pred = [0.86, 0.14, 1., 0., 1.]

from sklearn.metrics import precision_recall_fscore_support as score

precision, recall, fscore, support = score(y_test, y_pred)
print('precision: {}'.format(precision))
print('recall: {}'.format(recall))
print('fscore: {}'.format(fscore))
print('support: {}'.format(support))

So if you want to calculate these metrics you have to decide at some manner which values of the predicted vector are 1 (positive prediction) and which are 0 (negative prediction). For example, you can use a threshold (eg 0.5), or multiple thresholds and then select the best one or plot a curve with the different metrics at different threshold levels (eg 0.1 , 0.2, 0.3 and so on).