Proof-of-Concept of a Trust-based Recommender System
This project is related to our research lines: Systemic Risk and Financial networks Design and analysis of socio-technical systems
Duration 18 months (May 2008 - October 2009)
Funding source MTEC Foundation, Stiftung zur Förderung der Forschung und Ausbildung in Unternehmenswissenschaften an der ETH Zürich
Recommender Systems (RS) are applications that enable users of a particular online platform, e.g. Amazon, Last.FM, etc., to retrieve information on products and services offered. This information can be provided at different levels of personalisation and filtering. Thus, RS can be seen as tools to support the decision-making of consumers; because of this, they have become more and more widespread in all economic sectors. This project extends a novel type of electronic RS, developed at our chair, towards a real-world application. Differently from existing RS, the proposed system leverages the fact that users are part of a real social network and that they trust each other to different extents depending on the context. The main benefit of this approach is that it offers personalisation i.e. the recommendations is tailored to each individual user.
The main objective of the project is to prove the feasibility of a trust-based recommender system using social networks. We achieved this goal by developing a Knowledge Sharing Playground (KSP) which is, at a glance, a web application where users can share knowledge with other users. The framework implements the trust algorithms presented in Walter et al. (2006) and extended in Walter et al. (2009).
Currently, our group is negotiating a partnership with development companies in order to continue the development of the framework and to migrate it towards a commercial application. This will allow us to collect real data and validate empirically our models of RS.
Personalised and Dynamic Trust in Social Networks
|
[2009]
|
|
Walter, Frank Edward;
Battiston, Stefano;
Schweitzer, Frank
|
Proceedings of the third ACM conference on Recommender systems
pages: 197-204
|
more» «less
|
Abstract
We propose a novel trust metric for social networks which is suitable for application in recommender systems. It is personalised and dynamic and allows to compute the indirect trust between two agents which are not neighbours based on the direct trust between agents that are neighbours. In analogy to some personalised versions of PageRank, this metric makes use of the concept of feedback centrality and overcomes some of the limitations of other trust metrics.In particular, it does not neglect cycles and other patterns characterising social networks, as some other algorithms do. In order to apply the metric to recommender systems, we propose a way to make trust dynamic over time. We show by means of analytical approximations and computer simulations that the metric has the desired properties. Finally, we carry out an empirical validation on a dataset crawled from an Internet community and compare the performance of a recommender system using our metric to one using collaborative filtering.
Impact of Trust on the Performance of a Recommendation System in a Social Network
|
[2006]
|
|
Battiston, Stefano;
Walter, Frank Edward;
Schweitzer, Frank
|
Proceedings of the Workshop on Trust at the Fifth International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS'06)
|
more» «less
|
Abstract
Social agents naturally use their social and professional networks to filter information
by trustworthiness. In this paper, we present a model of an automated distributed recom-
mendation system on a social network and we investigate how the dynamics of trust among
agents affect the performance of the system. Agents search their social network for recom-
mendations on items to be consumed and the propagation of the query through agents at
several degrees of separation enhances the efficiency of their search. Moreover, agents have
heterogeneous preferences so that trust between neighbours can be used to filter information
coming from remote agents. We identify the range of the density of the network and the
degree of heterogeneity of agent preferences in which trust improves the performance of the
recommendation system.
|