November 2017
Beginner to intermediate
366 pages
7h 59m
English
Factor-based models leverage matrix decomposition techniques to predict recommendations for unrated items.
A model using funkSVD from recommenderlab is shown here:
> plan <- evaluationScheme(data, method="cross", train=0.9, given = 10, goodRating=5)> results <- evaluate(plan, method = "SVDF", type = "topNList", n = c(5,10,15) )SVDF run fold/sample [model time/prediction time] 1 [31.933sec/2.148sec] 2 [29.701sec/1.405sec] 3 [31.053sec/1.534sec] 4 [30.957sec/1.323sec] 5 [31.157sec/1.321sec] 6 [30.675sec/1.306sec] 7 [30.701sec/1.508sec] 8 [30.479sec/1.283sec] 9 [31.163sec/1.354sec] 10 [31.164sec/1.328sec] > avg(results) TP FP FN TN precision recall TPR FPR5 1.358667 3.641333 14.19933 70.80067 0.2717333 0.1047871 ...
Read now
Unlock full access