python - Pybrain gives wrong percentage of error -

i trained pybrain on huge dataset consisting of 3600 patterns , tried test in on similar dataset. percentage error shown 0.0277, predicted output array of 1s differs original output.

df = pandas.read_csv('bom-del.csv').groupby('x').mean().reset_index()[:3600] dk = pandas.read_csv('blr-del.csv').groupby('x').mean().reset_index() dx = numpy.array((dk.x-numpy.amin(dk.x))/(numpy.amax(dk.x)-numpy.amin(dk.x)) dy = numpy.array((dk.y-numpy.amin(dk.y))/(numpy.amax(dk.y)-numpy.amin(dk.y)) dx1 = numpy.array((df.x-numpy.amin(df.x))/(numpy.amax(df.x)-numpy.amin(df.x)) dy1 = numpy.array((df.y-numpy.amin(df.y))/(numpy.amax(df.y)-numpy.amin(df.y)) ds = classificationdataset(1, 1 , nb_classes=3600) dt = classificationdataset(1,1, nb_classes=3600) k in xrange(len(dx)):     ds.addsample(dx[k],dy[k])     dt.addsample(dx1[k],dy1[k]) trndata = dt tstdata = ds fnn = buildnetwork( trndata.indim, 360 , trndata.outdim, outclass=softmaxlayer ) trainer = backproptrainer( fnn, dataset=trndata, momentum=0.1, learningrate=0.01 , verbose=true, weightdecay=0.01)  trainer.trainepochs (50) print 'percent error on test dataset: ' , percenterror(     trainer.testonclassdata (        dataset=tstdata )        , tstdata['class'] )  numpy.array([fnn.activate(k) k,_  in tstdata]) 

i following results:

>>> array([[1.0],[1.0]....[1.0]) percentage error = 0.02 

although output should range between 0.001 1

can please explain why happening?