def classificationReport(self, doEvalSet=False):
if doEvalSet: #produce report for dev and eval sets instead
X_test = self.data.X_eval
y_test = self.data.y_eval
X_train = self.data.X_dev
y_train = self.data.y_dev
append = '_eval'
else:
X_test = self.data.X_test
y_test = self.data.y_test
X_train = self.data.X_train
y_train = self.data.y_train
append = ''
if not os.path.exists(self.output): os.makedirs(self.output)
f = open(
os.path.join(self.output,
'classificationReport' + append + '.txt'), 'w')
f.write('Performance on test set:')
classificationReport(self.bdt.predict(X_test),
self.bdt.decision_function(X_test), y_test, f)
f.write('\n')
f.write('Performance on training set:')
classificationReport(self.bdt.predict(X_train),
self.bdt.decision_function(X_train), y_train, f)
if self.crossValResults is not None:
f.write('\n\nCross Validation\n')
f.write("Cross val results: %.2f%% (%.2f%%)" %
(self.crossValResults.mean() * 100,
self.crossValResults.std() * 100))
def classificationReport(self, doEvalSet=False, batchSize=32):
if doEvalSet: #produce report for dev and eval sets instead
X_test = self.data.X_eval
y_test = self.data.y_eval
weights_test = self.data.weights_eval
X_train = self.data.X_dev
y_train = self.data.y_dev
weights_train = self.data.weights_dev
append = '_eval'
else:
X_test = self.data.X_test
y_test = self.data.y_test
weights_test = self.data.weights_test
X_train = self.data.X_train
y_train = self.data.y_train
weights_train = self.data.weights_train
append = ''
if not os.path.exists(self.output): os.makedirs(self.output)
f = open(
os.path.join(self.output,
'classificationReport' + append + '.txt'), 'w')
f.write('Performance on test set:\n')
report = self.model.evaluate(X_test.as_matrix(),
y_test.as_matrix(),
sample_weight=weights_test,
batch_size=batchSize)
self.score = report
if self.doRegression:
f.write('\n\nDNN Loss, Mean Squared Error:\n')
else:
classificationReport(
self.model.predict_classes(X_test.as_matrix()),
self.model.predict(X_test.as_matrix()), y_test, f)
f.write('\n\nDNN Loss, Accuracy, Significance:\n')
f.write(str(report))
f.write('\n\nPerformance on train set:\n')
report = self.model.evaluate(X_train.as_matrix(),
y_train.as_matrix(),
sample_weight=weights_train,
batch_size=batchSize)
if self.doRegression:
f.write('\n\nDNN Loss, Mean Squared Error:\n')
else:
classificationReport(
self.model.predict_classes(X_train.as_matrix()),
self.model.predict(X_train.as_matrix()), y_train, f)
f.write('\n\nDNN Loss, Accuracy, Significance:\n')
f.write(str(report))
if self.crossValResults is not None:
f.write('\n\nCross Validation\n')
f.write("Cross val results: %.2f%% (%.2f%%)" %
(self.crossValResults.mean() * 100,
self.crossValResults.std() * 100))