Need to figure out how to get more data. Will get one year more of data and apply argumentation to this to add even more data. Also, number of examples from each category should be equal ideally. Need to figure out a way for that. Train on 33000 samples
Epoch 1/3 33000/33000 [==============================] - 4s 113us/sample - loss: 3.9818 - accuracy: 0.7880 - sparse_top_k_categorical_accuracy: 0.9139
Epoch 2/3
33000/33000 [==============================] - 3s 87us/sample - loss: 0.6121 - accuracy: 0.8367 - sparse_top_k_categorical_accuracy: 0.9245
Epoch 3/3 33000/33000 [==============================] - 3s 88us/sample - loss: 0.5221 - accuracy: 0.8473 - sparse_top_k_categorical_accuracy: 0.9240