Can someone explain the importance of applying sklearn's train_test_split?

lmann2 wrote:

Thought this was helpful: https://www.researchgate.net/post/Whats_the_difference_between_training_set_and_test_set and http://stats.stackexchange.com/questions/19048/what-is-the-difference-between-test-set-and-validation-set .  My question still remains for this assignment though. 

Due to the possibility of overfitting, we cannot report accuracy using the same data used to train our classifier. We need to test accuracy on a fresh set of data: data that wasn't used in training.