Optimization Of Online Tour Operators' Personalized Travel Recommendation Based On Expander Graph

In recent years,tourism has continued to develop rapidly and has made a significant contribution to the economies of many nations around the world.Travel recommender systems are widely used to help tourists find interesting tourism products and services from excessive and unorganized information.Meanwhile,these systems also help online tour operators to make personalized travel recommendations so as to surviving in today's competitive market.Most existing recommendation techniques focus solely on tourist preferences and improving purchase probability.However,it has been increasingly recognized that the benefit of online tour operators under stochastic tourist demand is also an important aspect should be considered in tourism recommendations.In this article,we provide a focused study on how online tour operators manage and recommend items in the recommendation process,with the purpose of achieving low recommendation-related costs.The main research works in this paper are as follows:First,we investigate the state-of-the-art researches of travel recommendation,based on which we further indicate the shortcomings of present works and the importance of the work in this paper.Second,from the perspective of online tour operators,we formulate the tourism recommendation problem by taking the tourism recommendation-related costs and stochastic tourist demand into consideration.Third,we propose a expander graph based recommendation optimization model to optimize the top-N recommendation strategy,which focuses on tourists' preferences only.Moreover,a cutting plane procedure is put forward to solve this nonlinear discrete optimization problem.Finally,detailed experimental studies with simulated instances also conducted to verify the effectiveness of the expander graph based recommendation method,which demonstrate that high demand fill rate and low costs can be easily achieved by a expander graph structure.The paper also discusses the computational complexity,solution quality and robustness of the proposed method.