question about train_test_split()

[mpd]

It's just indicating that it's a long integer.

But the data I have are mix of negative and positive fraction integers, how can I avoid displaying the L?

Shape is just the number of samples multiplied by the number of features; it has nothing to do with what the data actually is.

This will print without the L:

print("test_y size = {0}".format(test_y.shape[0]))

Incidentally, Python 3 doesn't differentiate between int's and long's, so there is no 'L'. And unless you have a really good reason for using 2, you should be working with Python 3, anyway.