06/05/2014	Version 2.0 updated! Local collaborative ranking and rank-based SVD are available now!
06/13/2013	Version 1.2 updated! LLORMA is available now!
05/18/2012	A comparative study paper updated on arXiv.
04/20/2012	Version 1.1 updated!
07/06/2011	New algorithm added: Non-linear Matrix Factorization
06/01/2011	The website opened!

• Joonseok Lee, Mingxuan Sun, Guy Lebanon. PREA: Personalized Recommendation Algorithms Toolkit, Journal of Machine Learning Research (JMLR) 13:2699-2703, 2012. [BibTex]
• Joonseok Lee, Mingxuan Sun, Guy Lebanon. A Comparative Study of Collaborative Filtering Algorithms, ArXiv Report arXiv:1205.3193, 2012.

• A new algorithm LLORMA was added.
• Some bugs were fixed.

• Source code was refactored to reflect top-level design of class structure.
• A new baseline Overall Average was added.
• Input command interface was changed.
• Author guide and skeleton code for building custom recommendation algorithm is provided.
• Unit test module is added to help help developers.
• New dataset are added: Netflix small and MovieLens 1M.
• Performance benchmark is added to the Features page.
• Some bugs were fixed.

PREA (Personalized Recommendation Algorithms Toolkit) is an open source Java software that provides easy comparison of collaborative filtering algorithms. With increase demand of personalized services in e-commerce, recommendation systems are playing a critical role in commercial websites. In academia, many researchers have tried to achieve better performance and accuracy with various algorithms. Netflix Prize, held from 2006 to 2009, also contributed to take attention to research in collaborative filtering and recommendation systems. Our software provides a unique interface to compare several representative recommendation algorithms with common datasets as well as with your own dataset.

Our toolkit is aimed to the following users:

• A commercial merchandizer who wishes to build the best recommendation system for his or her website
• An algorithm designer who wants to test the performance of new recommendation algorithm in a standard way
• An academic researcher who wants to compare performance and accuracy of several recommendation algorithms in a fair and easy manner
• A software engineer who is interested in open source software and is willing to contribute to it

We are going to update the toolkit or this website continuously, including the following:

• Several more state-of-the-art recommendation algorithms
• New evaluation metrics, specially designed for commercial use
• Discussion section, aimed to receive your feedback