def accuracy_auc(y, ypred, weights=None):
"""Compute the accuracy, AUC, precision, recall and f1"""
from mozsci.evaluation import classification_error, auc_wmw_fast, precision_recall_f1
prf1 = precision_recall_f1(y, ypred, weights=weights)
return {'accuracy': 1.0 - classification_error(y, ypred, weights=weights),
'auc': auc_wmw_fast(y, ypred, weights=weights),
'precision': prf1[0],
'recall': prf1[1],
'f1': prf1[2]}
def accuracy_auc(y, ypred, weights=None):
"""Compute the accuracy, AUC, precision, recall and f1"""
from mozsci.evaluation import classification_error, auc_wmw_fast, precision_recall_f1
prf1 = precision_recall_f1(y, ypred, weights=weights)
return {'accuracy': 1.0 - classification_error(y, ypred, weights=weights),
'auc': auc_wmw_fast(y, ypred, weights=weights),
'precision': prf1[0],
'recall': prf1[1],
'f1': prf1[2]}
def test_precision_recall_f1(self):
tp = 1.0
fp = 3.0
fn = 2.0
actual_prec_rec_f1 = Test_precision_recall_f1.prec_rec_f1_from_tp_fp_fn(tp, fp, fn)
for y in [self.yactual, self.yactual1]:
for ypred in [self.ypred, self.ypred1]:
prec_rec_f1 = evaluation.precision_recall_f1(y, ypred)
for k in range(3):
self.assertTrue(abs(actual_prec_rec_f1[k] - prec_rec_f1[k]) < 1e-12)
def test_degenerate(self):
# test case with degenerate input
y = np.array([0])
ypred = np.array([[ 1.0]])
weights = np.array([1])
prf = evaluation.precision_recall_f1(y, ypred, weights=weights)
# check that none are NaN
self.assertFalse(np.array([np.isnan(ele) for ele in prf]).any())
# and they should all be 0
self.assertTrue(np.allclose(prf, [0, 0, 0]))
def test_degenerate(self):
# test case with degenerate input
y = np.array([0])
ypred = np.array([[1.0]])
weights = np.array([1])
prf = evaluation.precision_recall_f1(y, ypred, weights=weights)
# check that none are NaN
self.assertFalse(np.array([np.isnan(ele) for ele in prf]).any())
# and they should all be 0
self.assertTrue(np.allclose(prf, [0, 0, 0]))
def test_precision_recall_f1_weighted(self):
tp = 5.0
fp = 2.0 + 3 + 4
fn = 6.0 + 7
actual_prec_rec_f1 = Test_precision_recall_f1.prec_rec_f1_from_tp_fp_fn(tp, fp, fn)
for y in [self.yactual, self.yactual1]:
for ypred in [self.ypred, self.ypred1]:
for weights in [self.weights, self.weights1]:
prec_rec_f1 = evaluation.precision_recall_f1(y, ypred, weights=weights)
for k in range(3):
self.assertTrue(abs(actual_prec_rec_f1[k] - prec_rec_f1[k]) < 1e-12)
def test_precision_recall_f1(self):
tp = 1.0
fp = 3.0
fn = 2.0
actual_prec_rec_f1 = Test_precision_recall_f1.prec_rec_f1_from_tp_fp_fn(
tp, fp, fn)
for y in [self.yactual, self.yactual1]:
for ypred in [self.ypred, self.ypred1]:
prec_rec_f1 = evaluation.precision_recall_f1(y, ypred)
for k in xrange(3):
self.assertTrue(
abs(actual_prec_rec_f1[k] - prec_rec_f1[k]) < 1e-12)
def test_precision_recall_f1_weighted(self):
tp = 5.0
fp = 2.0 + 3 + 4
fn = 6.0 + 7
actual_prec_rec_f1 = Test_precision_recall_f1.prec_rec_f1_from_tp_fp_fn(
tp, fp, fn)
for y in [self.yactual, self.yactual1]:
for ypred in [self.ypred, self.ypred1]:
for weights in [self.weights, self.weights1]:
prec_rec_f1 = evaluation.precision_recall_f1(
y, ypred, weights=weights)
for k in xrange(3):
self.assertTrue(
abs(actual_prec_rec_f1[k] -
prec_rec_f1[k]) < 1e-12)