Easyrec is an open source recommender system that aims primarily for simple integration and administration. It ships as self-contained Web application with a graphical administration interface. User action tracking and delivery of recommendations are provided via REST or SOAP Web services. Recommendation algorithms are integrated through a plugin mechanism and can be added or removed at runtime.
The main idea of EasyRec is to provide an interface between existing recommender technology and integrators, focusing on the needs of those who administer the client application. Technically, EasyRec runs off-line recommender algorithms and stores the generated recommendations in a database, providing fast access to recommendations that can be arbitrarily complex to compute.
The admin application allows
- management of multiple client applications
- excluding specific items from recommendations
- reviewing recommendations
- statistics about the client application usage
- management of item types
- new in upcoming release: manual creation of item clusters e.g. for cross-selling
- managing recommendation algorithms (aka generator plugins)
Generator Plugin System
Generators plugins are wrappers around recommender implementations. They are executed on a regular basis and can be added by uploading a jar file in the admin interface. Currently shipping with easyrec:
- Association rule miner for calculating similar items based on user actions
- Slope-One for creating rating-based item recommendations