L.E.B.R.O.N.

• 

• 

• 

• 

  Hard at work on the best kind of whiteboard: ECSW windows

Inspiration

We were inspired by other NBA predictions, especially Nate Silver. Since Silver's models are purely algorithmic, we wanted to take it a step further and implement AI to better fine-tune out models. In addition, we were inspired by natural language processors like ChatGPT to make a natural language processor assistant.

What it does

It is a chatbot designed to predict the outcome of NBA games and analyze the predictions to give the user instructions to make favorable bets. In addition, the user is spoken to by none other than LeBron James himself.

How we built it

The model was built and trained on TensorFlow, and the chatbot was made using React. The integration was done using the fastapi library in python, and the front end was made using CSS.

Challenges we ran into

We had troubles implementing text to speech with a corresponding avatar that moves with the speech. In addition, we had troubles fine-tuning the regression and classification models at first.

Accomplishments that we're proud of

The chatbot pulls from multiple sources and is able to use multiple tools, including a custom tool, to talk, remember conversations, and make accurate predictions. In addition, our classification and regression models perform at just under industry standard with some training and fine tuning.

What we learned

We learned how to use python ML, react, lang chain, and how to work greatly as a team under strict time constraints.

What's next for L.E.B.R.O.N.

In the future, we would be able to improve the chatbot by improving the natural language processing, the face mapping for the avatar, the robustness of the models, and generalize to more sports.

Built With

• css
• fastapi
• langchain
• python
• react
• tensorflow