bookrater

Predicts user's book ratings based on a collaborative filtering model. Part of BookWeb

Installation

The recommended way is to use pipenv:

pipenv install --dev

There will be an error during locking, that's tolerable.

On Debian-based systems it's likely that you will also have to install Tkinter for Python:

# apt install python3-tk

Exploration

Start the app with

FLASK_APP=bookrater.py flask run

Then go to the localhost:5000/graphql, you will see the GraphiQL interface. You can make a query - GraphiQL gives you autocompletion and shows what types arguments should have. It also provides introspection into available operations and their arguments with a button in the top right.

Queries

You can query the model to get predicted ratings for pairs of the form (<user id>, <book id>). The server will return a list of predicted ratings. Queries have the following form:

{
    predictedRatings(users: <list of user ids>, books: <list of book ids>)
}

Mutations

You can retrain the model on new data.

mutation {
    retrain(users: <user ids>, books: <book ids>, ratings: <ratings>)
}