Research On Group Recommendation Method Of Ridesharing Based On Trust Relationship

Ridesharing is an important part of urban transportation,and the safety and trust experience of ridesharing passengers is an important factor in promoting multi-person ridesharing.Using social relationships existing in social networks to recommend ridesharing groups is an important method to improve ridesharing user experience.This method needs to integrate social relationship information and geographic location information,and faces the problem of sparse sharing trust information.Aiming at the above-mentioned problems,the thesis studies a group recommendation method for ridesharing based on trust relationship.The thesis uses Point Of Interest(POI)preferences and social relationships to model the trust relationship between ridesharing travelers,and transforms the problem of group ridesharing group recommendation based on the trust relationship into the integration and integration of multiple correlations.Based on this calculation,a group recommendation method for ridesharing based on trust relationship is constructed.The specific research content includes: First of all,for the diversity of relationships between travelers(such as travel relationships and trust relationships),a multi-correlation analysis and modeling method for travelers for ridesharing is proposed.This method constructs a single view and performs correlation coding for the temporal and spatial associations and social attributes of travelers.Among them,extract traveler pairs that meet travel time and space constraints to build a travel view,and mine traveler's social attribute data to build a trust view.Second,a multi-relevance ridesharing group recommendation algorithm(Multi-relevance ridesharing group recommendation algorithm,referred to as MRGR in the paper)is proposed.In this method,in order to integrate multiple correlations,the attention mechanism is expanded to collaboratively learn the above two types of view representations to obtain all traveler vectors and calculate the correlation of each pair of drivers and passengers.For each driver,sort and recommend passengers with high correlation to form a ridesharing group.Finally,the real location-based social data set Gowallo is used to test and verify MRGR.In the test,the group recommendation related parameters L,D,etc.are analyzed,and the parameter values with the best recommendation effect are obtained.At the same time,the evaluation indicators of the MRGR algorithm and other recommendation algorithms are compared item by item.The experimental results show that the proposed MRGR method is better than the existing methods in terms of network representation learning and group recommendation.The proposed MRGR method can more accurately recommend ridesharing groups with trust relationships.