Random Recommender Method

load recommenderlab

library("recommenderlab")

load data

data(Jester5k)

Data Properties

Jester5k

5000 x 100 rating matrix of class ‘realRatingMatrix’ with 362106 ratings.

Query registry for recommendation methods

recommenderRegistry$get_entries(dataType = "realRatingMatrix")

$IBCF_realRatingMatrix Recommender method: IBCF Description: Recommender based on item-based collaborative filtering (real data).

$POPULAR_realRatingMatrix Recommender method: POPULAR Description: Recommender based on item popularity (real data).

$RANDOM_realRatingMatrix Recommender method: RANDOM Description: Produce random recommendations (real ratings).

$UBCF_realRatingMatrix Recommender method: UBCF Description: Recommender based on user-based collaborative filtering (real data).

Create IBCF Recommender System

r <- Recommender(Jester5k, method = "RANDOM")

Describe Recommender

r Recommender of type ‘RANDOM’ for ‘ratingMatrix’ learned using 5000 users.

Get recommendation systems

names(getModel(r)) [1] "range"

Get top-N model

getModel(r)$topN NULL

Create top-5 list from IBCF recommender system for users 1001 and 1002

recom <- predict(r, Jester5k[1001:1002,], n=5)

Describe recommendations

recom Recommendations as ‘topNList’ with n = 5 for 2 users.

Show recommended items

as(recom, "list")

#[[1]] [1] "j83" "j75" "j37" "j96" "j22"

#[[2]] [1] "j99" "j30" "j94" "j76" "j83"

Evaluate recommender

Create evaluation scheme

4-fold cross-validation

scheme <- evaluationScheme(Jester5k, method="cross", k=4, given=3,

•                        goodRating=5) 
  

#Describe scheme

scheme

Evaluation scheme with 3 items given

Method: ‘cross-validation’ with 4 run(s). Good ratings: >=5.000000 Data set: 5000 x 100 rating matrix of class ‘realRatingMatrix’ with 362106 ratings.

Evaluate recommender method IBCF

Evaluate top-3 list

results <- evaluate(scheme, method="RANDOM", n=c(1,3,5,10,15,20))

RANDOM run

 1  [0.006sec/1.744sec] 
 2  [0.006sec/1.874sec] 
 3  [0.008sec/1.789sec] 
 4  [0.007sec/1.697sec] 

Display results

results

Evaluation results for 4 runs using method ‘RANDOM’.

View confusion matrix for first fold

getConfusionMatrix(results)[[1]]

n TP FP FN TN PP recall precision FPR TPR 1 0.1848 0.8152 17.1656 78.8344 1 0.01065105 0.1848000 0.01023483 0.01065105 3 0.5464 2.4536 16.8040 77.1960 3 0.03149207 0.1821333 0.03080493 0.03149207 5 0.8920 4.1080 16.4584 75.5416 5 0.05141092 0.1784000 0.05157590 0.05141092 10 1.8040 8.1960 15.5464 71.4536 10 0.10397455 0.1804000 0.10290071 0.10397455 15 2.6744 12.3256 14.6760 67.3240 15 0.15414054 0.1782933 0.15474780 0.15414054 20 3.5760 16.4240 13.7744 63.2256 20 0.20610476 0.1788000 0.20620317 0.20610476

Average results for 4-fold cross-validation

avg(results)

n TP FP FN TN PP recall precision FPR TPR 1 0.1922 0.8078 17.6240 78.3760 1 0.01078638 0.19220 0.01020150 0.01078638 3 0.5610 2.4390 17.2552 76.7448 3 0.03148779 0.18700 0.03080172 0.03148779 5 0.9334 4.0666 16.8828 75.1172 5 0.05238309 0.18668 0.05135608 0.05238309 10 1.8454 8.1546 15.9708 71.0292 10 0.10358198 0.18454 0.10298328 0.10358198 15 2.7510 12.2490 15.0652 66.9348 15 0.15440694 0.18340 0.15469058 0.15440694 20 3.6870 16.3130 14.1292 62.8708 20 0.20693633 0.18435 0.20601384 0.20693633

ROC curve plot

plot(results, annotate=TRUE)

Precision-recall plot

plot(results, "prec/rec", annotate=TRUE)