NFL Big Data Bowl 2024 Pipeline: Predicting Forced Fumbles

Project Overview

The NFL Big Data Bowl is an annual competition where participants analyze NFL data and create innovative solutions to various football-related problems. In this project, my team and I train a random forest classification model to predict when a tackle will result in a forced fumble based on field position, charge caracteristics (speed, acceleration, etc.), player stature (height, weight, etc), and more. The model achieved an accuracy of 98.61% with a low rate of false positives/false negatives.

The data manipulation and analysis pipeline was created using Dagster. The package management is performed with rye.

Repository Conents

setup_venv.sh: This bash script helps download and set up Rye.

src/assets.py: This python script contains all the assets used in the Dagster pipeline, including those for data retrieval, data manipulation, and model training.

test/test_assets.py: This python script contains a few tests our team wrote to ensure proper pipeline functionality.

model_results/random_forest_results.json: This json contains the model accuracy and confusion matrix.

Getting Started

Set up the virtual environment

1. In a directory of your choice, run git clone https://github.com/pkirti33/nfl_bigdata_pipeline.git
2. Open a VS Code window with this project and open a new terminal.
3. Run bash setup_venv.sh.
4. Run source .venv/bin/activate

Set up the Kaggle API

1. Follow the Installation and Authentication instructions here.
2. Within the nfl_bigdata_pipeline directory, create a new file called .env. In this file, type in:

KAGGLE_USERNAME = "your kaggle username"
KAGGLE_KEY = "your kaggle API key" 

Launch the Dagster tool

1. In the terminal window, run dagster dev

Authors:

Pranav Kirti, Will Pagliaro, Ryan Lo, Joseph Lee @ Washington University in St. Louis