홈 › Papers

Multi-Faceted Trust Based Recommendation System

Recommendation systems play a decisive role in the choices we make on the internet. They seek to tailor decisions to a user. This makes trust a very important factor in recommendation systems, since it is believed that users are similar to the people that they trust, and hence will make similar choices to those users. Trust and its effects on the choices people make have been widely studied in the context of collaborative recommendation systems. It is understood that trust is not a single-faceted entity but can vary contextually. Recent research in the domain of trust based recommendation systems has shown that taking into account the facets of trust greatly improves the quality of recommendations (Mauro et al., 2019; Fang et al., 2015). We propose a recommendation system that takes multiple facets of trust into account while looking at how suitable a product might be for a particular user. The architecture proposed for this Multi-Faceted Trust Based Recommender (MFTBR) allows for extensibility - new facets of trust can be added without much effort - and dynamicity - trust facets are not weighed arbitrarily. Instead, the weights are optimised for the best result via a neural network. The trust facets considered here are local, global and category-wise trust. MFTBR performs significantly better than basic collaborative filtering - U2UCF (C. Desrosiers, 2011), as well as some established models in the domain of social and trust based recommendation systems - MTR (Mauro et al., 2019) and SocialFD (Yu et al., 2017). Thus, our model provides a better approximation of real-life recommendations, taking into account not only the impact of trust on recommendation, but the context in which trust is established.

Code (0)

등록된 구현이 없습니다.

Tasks

Similar Papers 제목 키워드 기반