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

pkg <- 'recommenderlab'

source("https://raw.githubusercontent.com/mhahsler/pkg_helpers/main/pkg_helpers.R")
pkg_title(pkg)

Introduction

Provides a research infrastructure to develop and evaluate collaborative filtering recommender algorithms. This includes a sparse representation for user-item matrices, many popular algorithms, top-N recommendations, and cross-validation. The package supports rating (e.g., 1-5 stars) and unary (0-1) data sets.

pkg_usage(pkg)
pkg_citation(pkg, 2)

Supported algorithms

Recommender algorithm

• User-based collaborative filtering (UBCF)
• Item-based collaborative filtering (IBCF)
• SVD with column-mean imputation (SVD)
• Funk SVD (SVDF)
• Alternating Least Squares (ALS)
• Matrix factorization with LIBMF (LIBMF)
• Association rule-based recommender (AR)
• Popular items (POPULAR)
• Randomly chosen items for comparison (RANDOM)
• Re-recommend liked items (RERECOMMEND)
• Hybrid recommendations (HybridRecommender)

Recommender Evaluation

The framework supports given-n and all-but-x protocols with

• Train/test split
• Cross-validation
• Repeated bootstrap sampling

Available evaluation measures are

• Rating errors: MSE, RMSE, MAE
• Top-N recommendations: TPR/FPR (ROC), precision and recall

pkg_install(pkg)

Usage

Load the package and prepare a dataset (included in the package). The MovieLense data contains user ratings for movies on a 1 to 5 star scale. We only use here users with more than 100 ratings.

set.seed(1234)

library("recommenderlab")
data("MovieLense")

MovieLense100 <- MovieLense[rowCounts(MovieLense) > 100, ]
MovieLense100

Train a user-based collaborative filtering recommender using a small training set.

train <- MovieLense100[1:300]
rec <- Recommender(train, method = "UBCF")
rec

Create top-N recommendations for new users (users 301 and 302).

pre <- predict(rec, MovieLense100[301:302], n = 5)
pre

as(pre, "list")

Use a 10-fold cross-validation scheme to compare the top-N lists of several algorithms. Movies with 4 or more stars are considered a good recommendation. We plot true negative vs. true positive rate for top-N lists of different lengths.

scheme <- evaluationScheme(MovieLense100, method = "cross-validation", k = 10, 
  given = -5, goodRating = 4)
scheme

algorithms <- list(
  "random items" = list(name = "RANDOM", param = NULL),
  "popular items" = list(name = "POPULAR", param = NULL),
  "user-based CF" = list(name = "UBCF", param = list(nn = 3)),
  "item-based CF" = list(name = "IBCF", param = list(k = 100))
)

results <- evaluate(scheme, algorithms, type = "topNList",
  n=c(1, 3, 5, 10), progress = FALSE)

plot(results, annotate = 2, legend = "topleft")

Shiny App

A simple Shiny App running recommenderlab can be found at https://mhahsler-apps.shinyapps.io/Jester/ (source code).

References

• Michael Hahsler (2022) recommenderlab: An R framework for developing and testing recommendation algorithms. arXiv:2205.12371 [cs.IR]. DOI: 10.48550/arXiv.2205.12371.
• recommenderlab reference manual
• Suresh K. Gorakala and Michele Usuelli (2015) Building a Recommendation System with R (Packt Publishing) features the package recommenderlab.

mhahsler/recommenderlab documentation built on Aug. 30, 2024, 11:21 p.m.