findMultiplicYdots: Multiplicative Model for Collaborative Filtering

In matloff/rectools: Advanced Recommender System

Multiplicative Model for Collaborative Filtering

Description

Multiplicative model for binary Y, analogous to the usual additive model for general Y.

Usage

findsMultiplicYdots(ratingsIn) 
predict.MMmultiplic(multiplicObj, testSet) 

Arguments

ratingsIn	Input data frame. Within-row format is UserID, ItemID, Rating and optional covariates.
multiplicObj	An object of class 'MMmultiplic'.
testSet	A data frame consisting of cases to be predicted. Format is the same as ratingsIn, in the first two columns. Further columns, if any, are ignored.

Details

Note: This software assumes that user and item ID number are consecutive, starting with 1.

The basic model is

probability of a 1 rating = b0 + b1 * alpha * beta

where alpha and beta are effects for the given user and item, calculated as the mean Y values for the given user and given item, respectively.

Currently the code works only on the original data set, to predict the missing values. Covariates are not yet allowed.

Value

The function findMultiplicYdots returns an object of class 'MMmultiplic', consisting of the alpha and beta vectors, the vector (b0,b1) and the mean Y value nu.

The function predict.MMmultiplic returns a 2-column matrix, consisting of the estimated probabilities of 1 and the rounded version of the latter.

Author(s)

Norm Matloff

Examples

# lme4 data set, needs some prep
ivl <- InstEval
ivl <- ivl[,c(1,2,7)]
# convert from factors
ivl$s <- as.numeric(ivl$s)
ivl$d <- as.numeric(ivl$d)
ivl$y <- as.numeric(ivl$y)
# make it binary
ivl$y <- as.integer(ivl$y >= 4)
# run the training data
ydout <- trainMultiplic(ivl[,1:3]) 
# form a test set to illustrate prediction; make template first
testSet <- ivl[c(3,8),]
# say want to predict whether students 3 and 8 would rate instructor 12
# at least a 4
testSet[1,2] <- 12
testSet[2,2] <- 12
predict(ydout,testSet)

matloff/rectools documentation built on March 31, 2022, 12:09 p.m.