zaratsian/nfl_predictions

NFL Predictions (WebApp with PySpark)

Spark NFL Predictions

This project deploys a pyspark model that is used to predict the next NLF play. Specfically, there are two models and I trained on historical NFL Play-by-Play data. There's one model that predicts the expected yards gained or lost for a running play, and similarly a model that predicts yards gained or lost for a passing play. The model will output the expected yardage as well as which play (running or passing) should be executed.

To run:

NOTE: Before you begin, make sure you have Docker installed. This app is deployed and runs within a docker container.

    1) Download this repo
    2) ./build.sh
    3) ./run.sh
    4) Wait a minute or two, then open up your browser and go to http://localhost:14444

Bonus (for real-time scoring):
I have included an Apache NiFi template, called NFL_Spark_Model_Scoring.xml. This template can be loaded into Apache NiFi. Once loaded, it'll enable users to ingest, process, and score real-time play-by-play data against this REST-based Spark Web App. The incoming data stream may come from an API, a message bus (such as Kafka or JMS), or another real-time (or static) data source.