def test_logreg(model_fname,test_fname,output_fname=None, prob=True,acc=True):
"""
Predict labels for the test instances using the trained
model. If prob is set to True, then show class probabilities.
If acc is set to True and if the test instances have labels,
then we will predict accuracies for the test instances.
If an output_fname is specified we will write the predictions to
the file instead of writing to the terminal.
"""
pred = PREDICTOR()
pred.loadModel(model_fname)
testFile = SEQUENTIAL_FILE_READER(test_fname)
count = 0
E = EVALUATOR(pred.n)
if output_fname:
output = open(output_fname,"w")
else:
output = sys.stdout
for mv in testFile:
v = mv["vect"]
(lbl,prob) = pred.predictVect(v)
output.write("%d\t%s\n" % (lbl,str(prob)))
if pred.n == 2 and v.label == -1 :
trueLabel = 0
else:
trueLabel = v.label
if v.label is not None:
E.add(trueLabel,lbl)
count += 1
testFile.close()
if acc:
result = E.getMetrics()
E.show(result)
pass
def get_performance(self, fvects):
"""
Compute precision, recall and F-scores with the current
weight vector for the fvects using the EVALUATOR..
"""
E = EVALUATOR(self.n)
pred = PREDICTOR()
pred.loadWeights(self.w, self.bias, self.n)
for v in fvects:
(lbl, prob) = pred.predictVect(v)
E.add(v.label, lbl)
return E
def get_performance(self,fvects):
"""
Compute precision, recall and F-scores with the current
weight vector for the fvects using the EVALUATOR..
"""
E = EVALUATOR(self.n)
pred = PREDICTOR()
pred.loadWeights(self.w,self.bias,self.n)
for v in fvects:
(lbl,prob) = pred.predictVect(v)
E.add(v.label,lbl)
return E
def test_logreg(model_fname,
test_fname,
output_fname=None,
prob=True,
acc=True):
"""
Predict labels for the test instances using the trained
model. If prob is set to True, then show class probabilities.
If acc is set to True and if the test instances have labels,
then we will predict accuracies for the test instances.
If an output_fname is specified we will write the predictions to
the file instead of writing to the terminal.
"""
pred = PREDICTOR()
pred.loadModel(model_fname)
testFile = SEQUENTIAL_FILE_READER(test_fname)
count = 0
E = EVALUATOR(pred.n)
if output_fname:
output = open(output_fname, "w")
else:
output = sys.stdout
for mv in testFile:
v = mv["vect"]
(lbl, prob) = pred.predictVect(v)
output.write("%d\t%s\n" % (lbl, str(prob)))
if pred.n == 2 and v.label == -1:
trueLabel = 0
else:
trueLabel = v.label
if v.label is not None:
E.add(trueLabel, lbl)
count += 1
testFile.close()
if acc:
result = E.getMetrics()
E.show(result)
pass