Science Fair Pre-event Reflection
My Science Fair project is an innovation designed to help sport betters make smarter decisions, especially in the sport of basketball. The project uses a machine learning algorithm to recognize potential patterns between player stats and the result of a game.
To collaborate with my partner, we used discord for communication and used Google Docs and Google Drive to share important project files.
Jordan and I are designing a deep learning neural network and training it to recognize patterns in players stats and results. This innovation will be entirely done on a computer. To design and train the neural network, I am using Python in conjunction with the Tensorflow framework and Keras (a Python deep learning library). The type of neural network I will be aiming to create is a deep feedforward network.
I have already consulted multiple articles about the specifics of building a neural network. For example, this article is where I learned the basics of artificial neural networks.
Jordan and I are planning to consult our local basketball experts (Mr. Lloyd) in order to learn more about vital stats that might be indicative of a player’s effect on the outcome of the game and the player’s contribution in the game.
Image of the code for pre-processing of the player data. (CSV file manipulation with Python)
Post – Reflection
We used to same tools to collaborate as the start of the project. (Please check out the Pre-Reflection, 2nd paragraph)
Our research question for this project was “Is it possible to predict NBA results with Machine Learning.” I feel that we have fully answered a research question with a solid yes. Our program proved to be successful with an accuracy of over 65% in predicting results. This technology will benefit sport enthusiasts and help them make better decisions when gambling. Another application for this technology is to help team coaches improve their team’s weaknesses by adjusting the player stats and watch the difference in prediction.
In the end, we successfully implemented the solution we came up with at the start of the project. There was some difficulty adjusting the parameters of the program and there were many bugs in the first version of the neural network. In the end, all of the bugs were fixed and parameters were tuned.
We have consulted over 100 articles and forum threads about different issues we ran into during implementation. There were a few sources that seemed conflicting at first, but after more research, there were actually many different ways to solve the problem.
We have consulted Mr. Lloyd (our local basketball expert) on which data points have the most impact on the game overall. It was really informative and we gained valuable insight into the world of basketball statistics.
Our science fair trifold.
Sources:
“API Documentation.” TensorFlow, https://www.tensorflow.org/api_docs. Accessed 12 Apr. 2019.
“Build from Source.” TensorFlow, https://www.tensorflow.org/install/source. Accessed 12 Apr. 2019.
3.7.3 Documentation. https://docs.python.org/3/. Accessed 12 Apr. 2019.