RecSys

RecSys.jl is an implementation of the ALS-WR algorithm from "Yunhong Zhou, Dennis Wilkinson, Robert Schreiber and Rong Pan. Large-Scale Parallel Collaborative Filtering for the Netflix Prize. Proceedings of the 4th international conference on Algorithmic Aspects in Information and Management. Shanghai, China pp. 337-348, 2008"

• Install: Pkg.clone("https://github.com/abhijithch/RecSys.jl.git")
• Specify the training dataset in one of several ways:
  • Use delimited (CSV) file with columns: user_id, item_id, ratings. E.g.: trainingset = DlmFile("ratings.csv", ',', true).
  • Use a MAT file, specifying the file and entry name. E.g.: trainingset = MatFile("ratings.mat", "training")
  • Provide an implementation of FileSpec for any other format.
• Initialize: als = ALSWR(trainingset)
• Train: train(als, num_iterations, num_factors, lambda)
• Check model quality:
  • rmse(als) to check against training dataset
  • rmse(als, testdataset) to check against a test dataset
  • and repeat training with different parameters till satisfactory
• Save model: save(als, filename)
• Load model: als = load(filename)
• Get recommendations:
  • recommend(als, user_id) for an existing user
  • recommend(als, user_ratings) for a new/anonymous user

See examples for more details:

• last.fm
• netflix
• movielens