gorse is a recommender system engine consists of two parts and five packages.

<img width=320px src=”https://img.sine-x.com/packages.png”>

The application layer provides CLI tools for cross-validation, data import, database initialization, and recommendation service.

The package layer provides components to build a recommender system. Base data structures are included in the base package and basic computations are implemented in the float package. The core package consists of data loader, data splitter, evaluator and model selection tools. All recommendation models are in the tool package.

All data are stored in a SQL database. There are five tables in the initialized database:

<img width=420px src=”https://img.sine-x.com/database.png”>

All items are stored in the items table while all ratings are stored in the ratings table. Recommendation models load data from the ratings table and learn parameters. Then, top n recommendations are cached in the ratings table and top n neighbor items are cached in the neighbors table. Metadata are saved in the status table.

## For Server Users: Setup a Recommender System

## For Package Users: Hack on Recommender System

Contents

Table of Contents