HybridRecommender {recommenderlab}

Create a Hybrid Recommender

Description

Creates and combines recommendations using several recommender algorithms.

Usage

HybridRecommender(..., weights = NULL, aggregation_type = "sum")

Arguments

...

objects of class 'Recommender'.

weights

weights for the recommenders. The recommenders are equally weighted by default.

aggregation_type

How are the recommendations aggregated. Options are "sum", "min", and "max".

Details

The hybrid recommender is initialized with a set of pretrained Recommender objects. Typically, the algorithms are trained using the same training set. If different training sets are used, then, at least the training sets need to have the same items in the same order.

Alternatively, hybrid recommenders can be created using the regular Recommender() interface. Here method is set to HYBRID and parameter contains a list with recommenders and weights. recommenders are a list of recommender alorithms, where each algorithms is represented as a list with elements name (method of the recommender) and parameters (the algorithms parameters). This method can be used in evaluate()

For creating recommendations (predict), each recommender algorithm is used to create ratings. The individual ratings are combined using a weighted sum where missing ratings are ignored. Weights can be specified in weights.

Value

An object of class 'Recommender'.

See Also

Recommender

Examples

data("MovieLense")
MovieLense100 <- MovieLense[rowCounts(MovieLense) >100,]
train <- MovieLense100[1:100]
test <- MovieLense100[101:103]

## mix popular movies with a random recommendations for diversity and
## rerecommend some movies the user liked.
recom <- HybridRecommender(
  Recommender(train, method = "POPULAR"),
  Recommender(train, method = "RANDOM"),
  Recommender(train, method = "RERECOMMEND"),
  weights = c(.6, .1, .3)
  )

recom

getModel(recom)

as(predict(recom, test), "list")

## create a hybrid recommender using the regular Recommender interface.
## This is needed to use hybrid recommenders with evaluate().
recommenders <- list(
  RANDOM = list(name = "POPULAR", param = NULL),
  POPULAR = list(name = "RANDOM", param = NULL),
  RERECOMMEND = list(name = "RERECOMMEND", param = NULL)
)

weights <- c(.6, .1, .3)

recom <- Recommender(train, method = "HYBRID",
  parameter = list(recommenders = recommenders, weights = weights))
recom

as(predict(recom, test), "list")