Coherence and Inconsistencies in Rating Behavior - Estimating the Magic Barrier of Recommender Systems
Entity
UAM. Departamento de Ingeniería InformáticaPublisher
SpringerDate
2018-04-13Citation
10.1007/s11257-018-9202-0
User Modeling and User-Adapted Interaction 28 (2018): 97–125
ISSN
0924-1868 (print); 1573-1391 (online)DOI
10.1007/s11257-018-9202-0Funded by
This research was in part supported by the Spanish Ministry of Economy, Industry and Competitiveness (TIN2016-80630-P)Project
Gobierno de España. TIN2016-80630-PEditor's Version
https://doi.org/10.1007/s11257-018-9202-0Subjects
Recommender Systems; Evaluation; Benchmarking; Noise; Rating Coherence; Context; Evaluation Metrics; Magic Barrier; Ratings; User Behavior; InformáticaRights
© The Author(s) 2018Abstract
Recommender Systems have to deal with a wide variety of users
and user types that express their preferences in di erent ways. This di erence
in user behavior can have a profound impact on the performance of the recommender
system. Users receive better (or worse) recommendations depending on
the quantity and the quality of the information the system knows about them.
Speci cally, the inconsistencies in users' preferences impose a lower bound on
the error the system may achieve when predicting ratings for one particular
user { this is referred to as the magic barrier.
In this work, we present a mathematical characterization of the magic
barrier based on the assumption that user ratings are a icted with inconsistencies
{ noise. Furthermore, we propose a measure of the consistency of user
ratings (rating coherence) that predicts the performance of recommendation
methods. More speci cally, we show that user coherence is correlated with the
magic barrier; we exploit this correlation to discriminate between easy users
(those with a lower magic barrier) and di cult ones (those with a higher magic
barrier). We report experiments where the recommendation error for the more
coherent users is lower than that of the less coherent ones. We further validate
these results by using two public datasets, where the necessary data to identify
the magic barrier is not available, in which we obtain similar performance
improvements
Files in this item
Google Scholar:Said, Alan
-
Bellogin Kouki, Alejandro
This item appears in the following Collection(s)
Related items
Showing items related by title, author, creator and subject.