Information Technology Reference
In-Depth Information
considered eight nodes having the type Actor . All these nodes have a direct edge
to the Actor node The Hobbit (2012) . In addition, paths via the relationship
sets Misc , Producers , Writers , and Directors are taken into account. Edge
weights and edge labels are not shown in the explanation graph in order to keep the
explanation simple.
The disadvantages of memory-based recommenders are that these recommenders
do not provide strategies for handling inconsistencies and noise. Due to the run-
time complexity of memory-based recommenders only short paths can be taken into
account.
In a nutshell, memory-based recommender approaches should be used, if the data
change frequently, the data do not require a model-based cleanup, and if human-
readable explanations should be provided.
5.4.2 Model-Based Recommender
Real-world datasets are often huge, noisy, and sparse [ 26 , 27 ]. Thus, adequatemodels
are needed to extract the relevant information and to remove irrelevant data (e.g.,
noise).
The computation of the low-rank approximation of the graph's adjacency matrix
is a popular technique for extracting the dominant information from a graph [ 17 ]. The
approximation can be implemented by calculating the singular value decomposition
(SVD) of the adjacency matrix A of the graph G . In order to focus on the most
dominant information, the first k latent dimensions are taken into account.
misc
Jed Brophy
Mark Hadlow
Craig Hall
Andy Serkis
Eric Vespe
Katie Jackson
Actors
Andy Buckley
Jared Connon
Mathew Gordon
Cale Hetariki
Jamie Lawrence
Veronique Lawrence
Steve Old
Brenna Townshend
The Hobbit
An unexpected
Journey (2012)
King Kong
(2005)
Directors
Peter Jackson
Writers
Producers
Philippa Boyens
Peter Jackson
Fran Walsh
Philippa Boyens
Carolynne Cunningham
Peter Jackson
Fran Walsh
Fig. 5.3 The figure visualizes the explanation for a path-based recommendation (used in our movie
recommendation web application). The graph shows the nodes and the edges taken into account
when suggesting The Hobbit (2012) for the user input King Kong (2005)
Search WWH ::




Custom Search