Hybrid Bee Colony Trust Mechanism in Recommender System

Abstract

In the era of internet, there is a lot of information on the web. The information whether implicit or explicit is growing at an exponential rate, therefore perplexity in choosing the products and services has also soared up. Thus recommender system, an automated filtering mechanism, has been established to filter the information according to individual’s behavior and preferences and provide best and accurate suggestions. Recommendation can be in the form of verbal reviews, reviews about a movie or a book in internet and newspaper, surveys, travel guides etc. Collaborative filtering is one of the most popular and mature techniques in recommender system and evaluates items on the basis of opinions of other people. But in spite of its adeptness, it still suffers from problems such as cold start, sparsity, scalability and is susceptible to attacks like grey sheep, shilling attacks etc. Thus a solution has been propounded, which involves the collaboration of artificial bee colony, a swarm intelligence method, and trusted graph mechanism with that of collaborative filtering. Swarm intelligence is an artificial intelligence technique to study the behavior of insects in distributed systems while trust is a measure of reliability of user based on its preferences and behavior with distinct context at a particular time period. Former possess the adaptation, self organization and distribution properties while latter removes the fake recommendations and develops faith in the system. Thus the hybrid of these techniques increases the accuracy and robustness of recommendations while eliminating the attacks prevailing in existing systems. The proposed framework has been compared with other existing recommender system approaches with different parameters and validated by using dataset of movies available online