Research And Implementation Of Site Selection Using Semantic Trajectories

In recent years,with the improvement of people's awareness about the importance of location,site selection problem has become one of the hottest research topics.The most important point in site selection is to identify whether people in this area have relevant needs.At present,most existing works about the demand analysis of Point-ofInterest(POI)in a region are based on manual surveys or trajectory data analyses.However,traditional spatiotemporal trajectories only have geographic information,which can not accurately identify the visited POI at people's destination.With the proliferation of location sharing services and popularity of social platforms,the emergence of semantic trajectory overcomes the shortcoming of traditional trajectories that need to infer the category of POI at destination.However,there is no specific algorithm for mining regional POI's category demand and recommending location by utilizing semantic trajectories.Based on Foot Voting theory,people's cross-regional mobilities can be used to deduce their regional category demands.Frequent pattern mining can indirectly mine regional category demands based on this theory.However,because of the setting of frequent threshold,some category demands may be ignored and recursive search in frequent pattern mining is too time-consuming.In addition,researchers mainly focus on specific category and analyze specific features from trajectories,which is lack of generality of common demand discovery.To overcome the above shortcomings,according to Foot Voting,our paper uses semantic trajectories to construct a site selection system based on regional POI category demand mining,which can meet the requirements of high accuracy,efficiency and generality.In our paper,the system is divided into four components: social media data preprocessing,cross-regional pattern extraction,regional category demand identification and location recommendation.In social media data preprocessing,after aggregating check-ins to form semantic trajectories,trajectories are segmented based on the max time gap limit.The geographic space is divided into grid cells.Mapping checkins and POIs to the corresponding area,regional patterns are formed and characteristics of regional POI categories are analyzed.Based on some interesting observations,the cross-regional pattern extraction designs a fast and effective algorithm to extract crossregional patterns from the regional patterns set.The regional category demand identification module uses the cross-regional patterns and the characteristics of regional POI categories to identify people's demands.In this module,our paper designs a demand quantification model to measure the intensity of regional category demand.In addition,two identification algorithms are designed which are based on segmentation and connection respectively.In the location recommendation,the system recommends appropriate categories and regions for users according to the quantified intensity of regional category demand.Finally,a complete site selection system using semantic trajectories can be obtained by integrating all components in order.Our paper designs experiments to verify the efficiency and effectiveness of the integrated system.Experimental results on real world datasets reveal the advantages of our proposed algorithms and model.Our paper also enumerates recommendation cases to further illustrate the effectiveness of our system.