jolinxql / py-metric Goto Github PK

The Python module to common measurement methods.

$ sudo python setup.py install

or alternatively building module

$ python setup.py build

>>> from pymetric import AllMetrics
>>> A = [1, 0, 1]
>>> B = [1, 0, 0]
>>> AllMetrics.measure(A, B, 'rmse')
0.57735026918962573
>>> AllMetrics.measure(A, B, 'ap')
0.83333333333333326

The all_in_one interface of pymetric is AllMetrics.measure(y_truth, y_pred, method).

• y_truth : a list of int/float, or filename that contains a value each line.
• y_pred : a list of float, or filename that contains a value each line.
• method : several options as follows:

Group 1 is for regression task. Inputs of y_truth should be a list of floats.

• abs_error : AbsError.abs_error
• rmse : RMSE.rmse
• r2 : R2.r2
• ndcg : NDCG.ndcg
• cos : Cosine.cos

Group 2 is for classification task. Inputs of y_truth should be a list of 0-1 values.

• precision : PR.precision
• recall : PR.recall
• f1_score : PR.f1_score
• ap : MAP.ap
• auc : AUC.auc
• ndcg : NDCG.ndcg

• [AMS] (https://www.kaggle.com/c/higgs-boson/details/evaluation)

##Author

pymetric is developed and maintained by Zuotao Liu ([email protected]).