Prediction of FRC matches using TBA API and machine learning.
All python source files are command-line based and will need to be run from the command line/terminal (ex. on Windows: python create_dataset.py [arguments]).
First, you will need to create a dataset for the year's game by running create_dataset.py with the following usage:
python create_dataset.py out_filepath tba_api_key year
where out_filepath is a file path to an uncreated or already existing dataset file (with format .NPZ), tba_api_key is a valid The Blue Alliance key that you can create in your TBA Account Dashboard: https://www.thebluealliance.com/account.
Then, you can start predicting matches by using (currently) either the match_linear_regression or match_deep_learning predictors.
The arguments for both of these programs are the same:
python match_deep_learning.py dataset_filepath tba_api_key year event_name current_match match_type
python match_linear_regression.py dataset_filepath tba_api_key year event_name current_match match_type
where current_match is the number of the current match that you want to predict and match_type is either qm, ef, qf, sf, or f (see the api docs for more information).
The scripts will first output a visualization of the model which you will need to close and then the red and blue scores.
python create_dataset.py example_dataset/DeepSpace.npz YOUR_TBA_API_KEY 2019
python match_linear_regression.py example_dataset/DeepSpace.npz YOUR_TBA_API_KEY 2019 "Galileo Division" 1 f
python match_deep_learning.py example_dataset/DeepSpace.npz YOUR_TBA_API_KEY 2019 "Galileo Division" 1 f
Keras==2.2.4
matplotlib==2.2.2
numpy==1.14.5
requests==2.18.4
scikit-learn==0.20.2
scikit-plot==0.3.7
Simply download the source files and install the required prerequisites above.
Shown below are the inputs of the 2016 StrongHold game (Ranking Score, Auto, Scale/Challenge, Goals, and Defense) and their correlation with the number of points that the alliance earned).
