Preference surrogate-assisted interactive personalized evolutionary search algorithm based on user behaviors
With the rapid growth of the number of users on internet,a lot of user-generated contents(UGCs)has been generated,and there has been information overload.This paper makes full use of UGCs to build a user in-terest preference model,and proposes a preference surrogate-assisted interactive personalized evolutionary search algorithm based on user behaviors.Combing the interactive evolutionary computing,it helps users search for the items that meet their potential needs and interest preferences from a massive search space.By using interaction behaviors,ratings and item category information,a user preference perception model based on restricted Boltz-mann machine is constructed to extract the user preference features.From the perspective of evolutionary optimi-zation,a surrogate model based on the user preference and its evolutionary strategies is designed to generate new individuals with the user preference,and predict the fitness value of new individuals to guide the evolutionary op-timization process.Meanwhile,according to new UGCs and model management mechanism,these models are dy-namically updated to timely track the user preference for the personalized evolutionary search.Through a large number of experiments in the real-world datasets,the feasibility and effectiveness of the proposed algorithm are verified in dynamic personalized search and recommendation tasks.
万方数据
维普
