R/predict.R

In mhahsler/recommenderlab: Lab for Developing and Testing Recommender Algorithms

setMethod("predict", signature(object = "Recommender"),
  function(object,
    newdata,
    n = 10,
    data = NULL,
    type = "topNList",
    ...) {
    if (!is(newdata, "ratingMatrix") && !is(newdata, "numeric"))
      stop("newdata needs to be a subclass of class ratingMatrix or a numeric vector with user IDs!")

    object@predict(
      object@model,
      newdata,
      n = n,
      data = data,
      type = type,
      ...
    )
  })

### helper to return ratings
returnRatings <- function(ratings,
  newdata,
  type = c("topNList", "ratings", "ratingMatrix"),
  n,
  randomize = NULL,
  minRating = NA) {
  type <- match.arg(type)

  ratings <- as(ratings, "realRatingMatrix")
  ratings <- denormalize(ratings)

  if (type == "ratingMatrix") {
    ### replace with known ratings. Removed. We return the approximation by the algorithm instead.
    #nm <- as(denormalize(newdata), "matrix")
    #rm <- as(ratings, "matrix")
    #rm[!is.na(nm)] <- nm[!is.na(nm)]
    #return(as(rm, "realRatingMatrix"))
    return(ratings)
  }

  ratings <- removeKnownRatings(ratings, newdata)
  if (type == "ratings")
    return(ratings)

  getTopNLists(ratings, n, randomize = randomize, minRating = minRating)
}