Confusion Matrix showing results on opposite diagonal after Keras prediction

Question

I have trained my Keras model to do 3-class sentiment analysis with more than 200k tweets.

model = Sequential()
model.add(Dense(10, input_dim = features.shape[1],