Evaluation is based on categorical Accuracy. 
The evaluation obtained here is based on 30% of the test data uploaded.