Added a script used to classify actions (slides on it were presented at ML Forum opening meeting)
Added documentation on how to run the script
Testing with current data, the test sample still only correctly guesses the action ~87% of the time. I'm guessing we wanted higher than that. Throwing the result into the server should be easy, but it may slow it down a bit.
Testing with current data, the test sample still only correctly guesses the action ~87% of the time. I'm guessing we wanted higher than that. Throwing the result into the server should be easy, but it may slow it down a bit.