- evaluationScheme now drops users with too few ratings with a warning.
- evaluationScheme creation is now faster for realRatingMatrix.

- Fixed issues with ratingMatrix with missing dimnames.
- UBCF does now also work for users with fewer than n nearest neighbors.

- Preparations for changes in coercion for Matrix 1.4.2

- Fixed similarity() and dissimilarity() after changes for Cosine in package proxy (reported by Artur Gramacki).
- dropNA now always creates a dgCMatrix.

- calcPredictionAccuracy now works with negative values for given (all-but-x). A negative value produces an error with instructions.
- We require now proxy version >= 0.4-26 which fixed a conversion bug for cosine similarity.
- RECOM_AR now respects already know items (code provided by gregreich).
- evaluate: keepModel = TRUE now works (bug reported by gregreich).
- Recom_SVD: fixed issue with missing values set to zero (bug reported by jpbrooks@vcu.edu)

- Ratings of zero are now fully supported. We use .Machine$double.xmin to represent 0 in sparse matices. zapsmall() can be used to change them back to 0.
- topNList has now a method c() to combine multiple lists.
- RECOM_AR: Ratings are now equal to quality measure used for ranking.
- HYBRIDRECOMMENDER: add “max” and “min” aggregation.
- removeKnownRatings is now sparse.
- RECOM_RANDOM now has parameter range to specify the rating range.

- The MovieLense data set includes now also user meta information.

- getConfusionMatrix() is deprecated. Use getResults() instead.
- added an example for how to evaluate hybrid recommenders.
- calcPredicition now also reports N.
- calcPredicition now stores the list length for multiple top-N lists as a column called n in the result (instead of using rownames).

- UBCF for binary data: Fixed normalization for option weighted (reported by bhawwash).
- Fixed problems with less than k neighbors (reported by weiy6).
- Fixed incorrect description of comparisons in vignette.

- ratingMatrix gained method hasRatings.
- Recommender gained method “HYBRID” to create hybrid recommenders. Now hybrid recommenders can also be used in evaluate().
- similarity gained parameters min_matching and min_predictive.

- predict for Recommender RANDOM now uses the correct user ids in the prediction (reported by aliko-str).
- fixed weight bug in Recommender UBCF (reported by aliko-str).
- Recommender UBCF now removes self-matches if item ids are specified in newdata. Specifying data in predict is no longer necessary. (reported by aliko-str).
- HybridRecommender now handles NAs in predictions correctly (was handled as 0).

- predict with type “ratingMatrix” now returns predictions for the known ratings instead of replacing them with the known values.
- Recommender methods Popular, AR and RERECOMMENDER now also return ratings for binary data (and thus can be used for HybridRecommender).
- Added a LIBMF-based recommender.

- evaluationScheme with negative numbers for given (all-but-x scheme) now works even if there are no given items left (reported by philippschmalen).

- Fixed bug in denormalization by column with z-score (reported by jackyrx).
- Fixed bug in predict with type “ratingMatrix” where known values were not denormalized (reported by MounirHader).

- Fixed bug in ALS_implicit (reported by equalise).
- getData for binaryRatingMatrix data with type “known” and “unknown” preserves now user ids/rownames (reported by Kasia Kulma).
- predict for HybridRecommender now retains user IDs (reported by homodigitus).
- Removed warning about using drop in subsetting ratingMatrices (reported by donnydongchen).

- predict for IBCF now returns top-N lists correctly.
- (cross) dissimilarity for binary data now returns the correct data type (reported by inkrement).

- Added recommender method ALS and ALS_implicit based on latent factors and alternating least squares (contributed by Bregt Verreet).
- Changes in recommendation method AR: Default for maxlen is now 3 to find more specific rules. Parameters measure and decreasing for sorting the rule base are now called sort_measure and sort_decreasing. New parameter apriori_control can be used to pass a control list to apriori in arules.
- The registry now has a reference field.

- Fixed bug in method IBCF with n being ignored in predict (reported by Giorgio Alfredo Spedicato).

- Added recommender RERECOMMEND to recommend highly rated items again (e.g., movies to watch again).
- Added a hybrid recommender (HybridRecommender).
- realRatingMatrix supports now subset assignment with [.
- RECOM_POPULAR now shows the parameters in the registry.
- RECOM_RANDOM produced now random ratings from the estimated distribution of the available recommendations (from a normal distribution with the user’s means and standard deviation).
- predict now checks if newdata (number of items) is compatible with
the model.

- getTopNLists and bestN gained a randomized argument to increase prediction diversity.
- Added getRatings method for topNList.

- FIX: rownames of newdata are now preserved in prediction output.
- We use testthat now.
- Normalization now can be done on rows and columns at the same time.
- SVD with column-mean imputation now folds in new users.
- Added Funk SVD (funkSVD and recommender SVDF).
- Added function error measures: MAE, MSE, RMSE, frobenius (norm).
- Jester5k contains now the jokes.
- MovieLense contains now movie meta information.
- topNLists now also contains ratings.
- Removed obsolete PCA-based recommender.

- Fixed several problems in the vignette.
- predict for realRatingMatrix accepts now type = “ratingMatrix” to returns a completed rating matrix.
- Negative values for given in evaluationScheme implement all-but-given evaluation.
- Method “SVD” used now EM-based approximation from package bcv.

- NAMESPACE now imports non standard R packages.

- Fixed NAMESPACE problems.
- Evaluation of ratings is now better integrated into evaluate.
- binarize keeps now dimnames.

- Many.