Time-Sensitive Travel Route Recommendation Method Based On Dynamic Transfer Graph

With the rapid development of mobile Internet and the extensive use of mobile devices,the social software of Gowalla and Foursquare provides the function of sharing and check-in and collects a large amount of check-in data that can reflect the travel rules of users.These data can effectively support travel route recommendation and search.The existing travel route recommendation methods that are based on check-in data do not consider that the popularity of attractions changes according to time and that the changing rule of popularity is also not invariable.The popularity of attractions may change according to the change of seasons,months,or even holidays.The characteristics of popularity change of China's Changbai Mountain are the best proof of the change according to time.The end of every summer and the beginning of every fall are the most popular periods due to the highest probability of Tianchi being visible.Ski peak in December will bring another popular peak.And popularity of Changbai Mountain in other time periods is at a low point.This paper puts forward method of time-sensitive travel route recommendation based on dynamic transfer and studies travel route recommendation with the variation of attraction popularity as a result of time change.The paper makes a quantitative representation according to time series about the change of attraction popularity within the region to recommend travel routes that are most suitable for users' travel time.Route recommendation of dynamic transfer graph can be used for the research on travel route recommendation,scenic spot construction and safety facility application.Due to the randomness of check-in locations of objects,obtaining and learning reasonable stable pattern rules as well as classifying these rules to get transfer graph pattern becomes a challenge to the paper.This paper proposes a pattern-building method of dynamic transfer graph based on hierarchy clustering,designs methods to remove outliers with popularity sequence and constructs stable pattern rules based on generalized processing,serialization and modularized processing,which effectively removes local anomalies as well as modularized automatic aggregation.The paper uses greedy strategy route recommendation algorithm for optimal route search.The search results are compared with other three methods from three aspects,the user return,model establishment efficiency and route recommendation efficiency.The three methods are respectively Month time-sensitive route recommendation,Season time-sensitive route recommendation and Initial preference route recommendation.The dynamic transfer graph model is built through collecting real data.Through full experiment,it can be verified that under different parameters,the profit of users increases by more than 10%compared with the existing work,which verifies accuracy and effectiveness of the proposed method.The efficiency of the recommended route increases by more than 20%,which shows that the proposed method is effective.