zoo_recommender_system

A Zooniverse projects recommendation system for individual Zooniverse users - wrapped in a JSON API.

Installation

Use Docker and Docker Compose to install dependencies, alternatively use your local R install to get the system running.

1. Run docker-compose build to build the recommender API container

2. Run docker-compose up to serve the recommender API on the localhost

• requires the ratings matrix and project dataset to have been created (see below)

Usage

To prepare the data for the recommender engine

1. Get a bash console for using R scripts docker-compose run --rm --service-ports --entrypoint bash recommender-api

2. Run the code to generate the model inputs Rscript ./R/binary_rating_analysis.R

• This sources the R script "./R/load_prep_data.R" which reads in three datasets and formats them for analyses:
  • 2021_user_classification_counts.csv
  • 2021_user_first_last_classification.csv
  • launch_approved_projects_20-Dec-2021.txt
• The script also saves the .rds datafiles needed for analyses/recommendation in the input_data folder

To run the recommender api

1. Run docker-compose up

Alternatively from within the docker bash shell run Rscript ./R/run.R

Then can use the swagger interface to query the API.

Endpoints

• GET /recommend?user_id=10

Fetch the recommeded Zooniverse projects for user with ID 10

curl http://localhost:8000/recommend?user_id=10