A complete open source recommender system for Shuup.
The recommener system provides a powerful API to use the recommendations anywhere, be it in admin, frontend templates or also through REST APIs.
WIP. For now, see unit tests for how to use this API.
Here is the list of features to be implemented in the first release:
- Popularity: a simple rank of most seen and sold products.
- Who bought this, also bought: suggest products based on product purchase history of other customers using collaborative filtering. Features like age, gender etc might be useful when ranking the products. User-based filtering.
- Who saw this, also saw: suggest products based on product visit history of other customers using collaborative filtering. Features like age, gender etc might be useful when ranking the products. Item-based filtering.
- Category similarity: suggest categories to be used in products based on the product name and description. Used especially in admin to help with product creation task. Content-based filtering.
- Product similarity: suggest products that are similar based on product images and/or name/description/categories. Can be used in frontend and also in admin. Content-based filtering.
Use Jupyter notebooks to create the models before adding methods into the API.
To create or modify noteboks, install dev requirements: pip install -r dev-requirements.txt
Then run it using notebooks path using this command at project root path: ./scripts/run-notebook
Apache 2.0