I am considering two types of systems - which might have more appropriate names:

Recommender systems: These recommender systems are based on collaborative filtering methods, both model-based and memory-based. They take a partially filled user/item matrix of ratings of items by users, and try to predict the missing ratings.

Trust systems: These systems take a filled user/item matrix of ratings, and try to detect a trustworthiness for each user - which can then be used to give more fair ratings. These systems could for example be used to eliminate overly negative users or fake ratings.

There seems to be a lot of overlap between these two types of systems. My question is, if it is possible to simulate one system using the other. For example - to simulate a trust system with a recommender system - one could remove a rating $r$, let the recommender system predict it, and see how big a difference the actual and predicted value have.

Is it possible to use such a trick? What is the correct terminology for these systems?

I am most interested in systems of the second type, but have so far stumbled upon most research papers on the first type.


Some references that might interest you:
1] Authoritative sources in a hyperlinked environment
Jon Kleinberg
JACM 1999 http://www.cs.cornell.edu/home/kleinber/auth.pdf

2] Truth discovery with multiple conflicting information providers on the web
Xiaoxin Yin, Jiawei Han, Philip Yu
KDD 2007 http://www.cs.uiuc.edu/~hanj/pdf/kdd07_xyin.pdf

3] A Bayesian approach to discovering truth from conflicting sources for data integration
Bo Zhao, Benjamin Rubinstein, Jim Gemmel, Jiawei Han
VLDB 2012 http://vldb.org/pvldb/vol5/p550_bozhao_vldb2012.pdf

If you check the references and look at papers that cite these papers, you should get lot more papers along these lines.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.