Research On Group Trip Recommendation Based On Check-in Data

With the development of mobile devices and location-based service,LBSNs blossom quickly.More and more LBSNs websites,like Gowalla?Foursquare?Flicker,allow user to record and share their location information,a variety of application in LBSNs bring mass data,these data contains rich knowledge,not only records users' action information in the virtual network,but also records users' action in the geographical world.By analyze the mass of data,we can find the users' behavior pattern in the geographical world and social network,learn users' preference.Then we can provide valuable guidance for the users.Trip recommendation is a significance research area in LBSNs.When users go out and travel,users can get travel route guidance from trip recommendation system by the phone.The research content is how to mine the large scale data in LBSNs and how to recommendation famous travel route to users.The existing studies of trip recommendation usually addressed the case of user's traveling alone to meet the demand of individual user in the recommended route as much as possible.Travel together is a common phenomenon,the research on group trip recommendation has the vital significance.For this requirement,this paper proposes a group trip recommendation.Group trip recommendation is differ from trip recommendation for a single user because it needs to think about each user's preference in the group.How to coordinate the demand of each user in the group is a vital issue in group trip recommendation.The target of group trip recommendation is to get a route which can guarantee the overall group high satisfaction and individuals' satisfaction small difference.The overall group high satisfaction refers to the users as a whole,the group satisfaction high and individuals' satisfaction small difference refers to the results of the recommendation is fairness to each user.The recommendation fabricated for group users is called group recommendation system.In general,GRS firstly integrate users' preferences or user' rating to items and formulate group model according different aggregate strategies.Then the system recommends items according to group model.The existing aggregates mainly include average strategy,average without misery and least misery.This paper proposes a Dynamic Aggregation Preference(DAP)strategy.According to the degree of satisfaction for current individual,DAP adjusts the group preference model dynamically that can guarantee the overall group high satisfaction and individuals'satisfaction small difference.Based on DAP strategy,constructing the route evaluation model that grades the satisfaction of route and returning the route with the highest score.The paper gives the framework of group trip recommendation.Firstly,the paper gives the definition of group trip recommendation issue,maps the route recommendation to the graph and converts trip recommendation issue to optimal route search in the graph.To get all routes meet to user's query and find one route with the max score,we propose two trip search algorithm.They are DFRS algorithm based on depth-first search and GSRS algorithm based on greedy strategy.In the experiment,the paper adopts the real website dataset in Gowalla and Foursquare.In the first part we compare the effectiveness of DAP aggregate strategy with other aggregate strategy from availability of the recommended and verify the effectiveness of DAP aggregate strategy.In the second part,we compare the advantages and disadvantages of DFRS algorithm and GSRS algorithm from effectiveness and efficiency.