epl-ml-mono-repo
Experimental Repo epl-ml-mono-repo
- EPL English Premier League
- ML Machine Learning
- mono-repo Lerna based JS mono repo
This is an experimental project intended to test the following technologies
- tensorflow
- typescript
- lerna
- greenkeeper/renovate/regenerate
The intention for this project is to overcome the learning required to build a predictive engine from EPL resuls win, loose and draw over the lifetime of the EPL in order to make an engine to predict future games.
As with all academic exercises it is hoped that the experiences gained from the testing and use of this may be applied in a general way to other data sets so that a prediction may be made.
requirements
Mongo
To run the integration tests a mongo db is required. Mongo is to be run on port 27107. Although this demonstration program could be run from a JSON file a decision to use mongo will allow the development of the model.
Installation
Lerna Installation
If lerna is installed globally run lerna bootstrap in the root folder. Otherwise do you can install lerna globally yourself.
npm i -g lernaor carry out an npm install in the root folder and lerna will install, then export the node_modules .bin folder into your PATH.
npm i
export PATH=$PATH:./node_modules/.bin
Build
Build the packages with the following command
lerna run buildHost the model
It will be necessary to run the model host as it provides the end points for the client and the client host which hosts the frontend
goto the packages epl-host-model folder and run the program to host the model.
cd packages/epl-host-model
npm run start
This will start a server that runs on port 3000 that serves the TensorFlow model and weights file.
Run a front end prdection
goto the epl-client folder.
cd packages epl-client
npm run start
Bring up the front end
The front end may be viewed on http://localhost:1234. This is the default for parcel based client applications.