RemixAutoML (version 0.11.0)

AutoRecommender: Automatically build the best recommender model among models available.

Description

This function returns the winning model that you pass onto AutoRecommenderScoring

Usage

AutoRecommender(data, Partition = "Split", KFolds = 1, Ratio = 0.75,
  Given = 1, RatingType = "TopN", RatingsKeep = 20,
  SkipModels = "AssociationRules", ModelMetric = "TPR")

Arguments

data

This is your BinaryRatingsMatrix. See function RecomDataCreate

Partition

Choose from "split", "cross-validation", "bootstrap". See evaluationScheme in recommenderlab for details.

KFolds

Choose 1 for traditional train and test. Choose greater than 1 for the number of cross validations

Ratio

The ratio for train and test. E.g. 0.75 for 75 percent data allocated to training

Given

The number of products you would like to evaluate. Negative values implement all-but schemes.

RatingType

Choose from "TopN", "ratings", "ratingMatrix"

RatingsKeep

The total ratings you wish to return. Default is 20.

SkipModels

AssociationRules runs the slowest and may crash your system. Choose from: "AssociationRules","ItemBasedCF","UserBasedCF","PopularItems","RandomItems"

ModelMetric

Choose from "Precision", "Recall", "TPR", or "FPR"

Value

The winning model used for scoring in the AutoRecommenderScoring function

See Also

Other Marketing Modeling: AutoMarketBasketModel, AutoRecomDataCreate, AutoRecommenderScoring

Examples

# NOT RUN {
WinningModel <- AutoRecommender(RatingsMatrix,
                                Partition = "Split",
                                KFolds = 1,
                                Ratio = 0.75,
                                Given = 1,
                                RatingType = "TopN",
                                RatingsKeep = 20,
                                SkipModels = "AssociationRules",
                                ModelMetric = "TPR")
# }