Skip to content

YujunZhong/hockey-game-prediction

Repository files navigation

Hockey Game Prediction

This repository includes three parts for hockey game prediction: model deployment, live workflows and interactive visualization. The training code for the models and the visualization of the NHL data are in other branches.

The project originally came from the Data Science class at the University of Montreal, and the work was done by Yujun Zhong, Tejas Shivakumar Kasetty and Kasra Laamerad.

Setup & Deployment Guide

Applications to run:

  • Serving (Flask)
  • Streamlit

Pre-requisite

  • Users should have docker installed in their machine.
  • Store your Comet ML API key in COMET_API_KEY environment variable.

Both these applications are can be easily deployed using dockers.

Method 1:

  1. Execute build.sh (shell script) to build the docker images of serving and streamlit application.

  2. Execute run.sh (shell script) to run the serving and streamlit containers using their respective docker images.

Method 2:

  • Run compose.sh (shell script) to build and run docker containers. This a single step process which uses docker-compose. Users can configure PORT and other variables via compose.env.

Default configuration:

serving   - 8890(PORT) 
streamlit - 8880(PORT)

Usage

Users can access the application using these URLs: localhost:8880 or 127.0.0.1:8880