Research On Mining And Recommendation Of Passenger Search Hotspots Based On Taxi Trajectory

In the large-scale taxi history tracks,there is a lot of hidden search strategy information of taxi passengers,which is the collective wisdom of taxi drivers.In recent years,more and more researchers have been interested in how to improve the revenue of drivers by digging out efficient passenger search strategies.However,most of these researches are simply based on the data statistics analysis of the original data,unable to solve the cold-start problem and hard to make full use of the relevant factors hidden in the data,such as the driver's experience in finding passengers.Based on the above problems,this paper proposes a PIAS(Passenger Induced Area Search)algorithm for location recommendation service.The algorithm can take full advantage of the relationship between the search areas and the driver's experience to dig out the group of taxi drivers with high passenger searching efficiency and the area with high passenger searching value.The main work is as follows:The first is the clustering analysis of the search areas.This paper proposes a FGP-DBSCAN algorithm based on fuzzy grid partition,which can quickly cluster the data of large-scale pickup points and analyze the spatial distribution of each search area in cities.Compared with the traditional P-DBSCAN(based on parallel density clustering algorithm of clear grid partition)algorithm,it can effectively solve the problem of too many clusters caused by the hard boundary of the grid.Secondly,based on HITS(Hyperlink-Induced Topic Search),a PIAS algorithm for locationoriented recommendation service and SAH-Tree(Search Area Hierarchical Tree)data structure are proposed.Among them,PIAS algorithm can make full use of the relationship between the search areas and the driver's experience to dig out the accurate value ranking of search areas.SAH-Tree data structure improves the efficiency of KNN algorithm in searching the areas by constructing the edge relation between the boundary nodes of each region at each level.Then,recommending a personalized passenger searching route for drivers.Firstly,an optimal passenger searching sequence will be planned for the current no-load drivers based on the group wisdom of drivers above mentioned.Then,the road attributes of each connecting section of the passenger searching sequence are modeled according to the taxi track data.In the end,on the basis of the driver's preference,a personalized passenger search route will be provided.Finally,more than 1600 taxi drivers' real track data in Hangzhou are used as the data set to carry out a large number of verification experiments on the proposed algorithm.The experimental results show that PIAS algorithm can dig out more accurate value ranking of search areas.Therefore,the KNN algorithm based on SAH-Tree structure also has a better search efficiency.