Controlled Expansion of Neighborhood in Trust Based Recommender Systems

A technique to identify good recommenders and make them trustworthy neighbours is presented. By having good recommenders in direct association, an agent will have an improved set of recommendations in lesser amount of time and computations as compared to getting recommendations transitively from neighbour of a neighbour and so on. While estimating trust on an unknown agent, the upper limit on the number of hops one needs to explore to reach that agent is also proposed, hence further minimizing time and computation complexity. Results of experiments conducted on a real dataset illustrate the efficiency and effectiveness of the proposed method.