# Take max value in preds rows as classification
# pred = np.zeros((len(X_test)))
# yint = np.zeros((len(X_test)))
# for row in np.arange(0,len(predictions)) :
# pred[row] = np.argmax(predictions[row])
# yint[row] = np.argmax(y_test[row])
# print("Classifier Accuracy = %d"%(metrics.accuracy_score(yint,pred)))
#####
# now to test
#####
print("Testing classifier on Test data")
print("Re-train with full training set")
X, y, unique_cuisines,classes,test_indices,Xtest = getdata(ngram_range=(1,2)) # import the data
clf2 = nnk(X,unique_cuisines,lr=0.1)
f = clf2.fit(X.toarray(), y.toarray(), nb_epoch=25, batch_size=1000,
validation_split=0.15, show_accuracy=True)
predictions = clf2.predict(X.toarray(), batch_size=1000, verbose=1)
writestackgen(predictions,'StackGen.DNN.1-2grams.adadelta.train.csv')
predictions = clf2.predict(Xtest.toarray(), batch_size=1000, verbose=1)
writestackgen(predictions,'StackGen.DNN.1-2grams.adadelta.test.csv')
# # Take max value in preds rows as classification
# pred = np.zeros((len(Xtest)))
# for row in np.arange(0,len(predictions)) :
# pred[row] = np.argmax(predictions[row])
if __name__ == '__main__':
# Do the training
res1 = np.genfromtxt('StackGen.DNN.1-2grams.train.csv',delimiter=',')
res2 = np.genfromtxt('StackGen.DNN.1-2grams.adadelta.train.csv',delimiter=',')
res3 = np.genfromtxt('StackGen.DNN.1grams.train.csv',delimiter=',')
res4 = np.genfromtxt('StackGen.NN.1-2grams.train.csv',delimiter=',')
res5 = np.genfromtxt('StackGen.NN.1grams.train.csv',delimiter=',')
trainmat = np.mean(np.array([res1,res2,res3,res4,res5]),axis=0)
# now normalize by row
# trainmatnorm = preprocessing.normalize(trainmat)
# pull in true values
_, y, _,_,_,_ = getdata(ngram_range=(1,2))
# Instantiate neural network
cfr = nnk(trainmat,y,lr=0.1)
# train it
cfr.fit(trainmat, y.toarray(), nb_epoch=25, shuffle=True,
batch_size=1000, validation_split=0.15,
show_accuracy=True, verbose=1)
# now get test values
tst1 = np.genfromtxt('StackGen.DNN.1-2grams.test.csv',delimiter=',')
tst2 = np.genfromtxt('StackGen.DNN.1-2grams.adadelta.test.csv',delimiter=',')
tst3 = np.genfromtxt('StackGen.DNN.1grams.test.csv',delimiter=',')
tst4 = np.genfromtxt('StackGen.NN.1-2grams.test.csv',delimiter=',')
tst5 = np.genfromtxt('StackGen.NN.1grams.test.csv',delimiter=',')
