Research On The Combined Recommendation Model Of Taxi Cruising Routes Based On GPS Data

Taxi is an important part of the urban transportation system and an indispensable means of transport for residents,plays an important role in meeting their short-distance and personalised travel needs.However,in the process of taxi operation,they often face the problem of finding passengers with empty loads.With the popularity of the online taxi operation mode,taxis have improved their operational efficiency through the online booking hailing mode,but this does not fundamentally solve the problems of their inefficient blind cruising;taking up road resources as well as difficulties in improving drivers’ income.Therefore,it is necessary to propose a reasonable cruising route recommendation method to assist taxi drivers in making cruising decisions,so as to improve the operational efficiency of taxis.Firstly,the paper processes and analyses the GPS data of taxis in Chengdu,and achieves the matching between the trajectory data and the actual road network map through the improved nearest neighbour matching method,and obtains the taxi pick-up and drop-off data and individual travel trajectory data from the trajectory data.The operational characteristics of taxis were quantified in terms of both temporal and spatial characteristics;secondly,the income of drivers was classified according to the operational efficiency of taxis and the operational characteristics of different types of drivers were compared.The results of the study show that taxi drivers have distinct spatial and temporal characteristics in the morning and evening peaks,and in the comparative analysis,the high-income driver group mainly obtains higher revenue through shorter cruising time and more efficient orders.Then,a recommendation model for taxi cruising route combinations is constructed,the hotspots of taxis are obtained through the density-based DBSCAN clustering method and divided into regions,taking into account the shortest route travel time,route probability of carrying passengers,taxi supply/demand ratio and expected trip revenue as the attribute indicators of the region,and then determining the attribute value weights according to the information entropy principle to construct the best recommendation model for cruising regions.Finally,based on the taxi journey time,the cruising route is recommended for the driver when he arrives at the target area,and the regional section recommendation model is established considering the traffic condition of the section and the probability of passenger loading,and the continuous optimal attribute section output is recommended for the driver as the cruising route using the breadth-first search strategy.Finally,an arbitrary drop-off point on a weekday is selected as the starting point of the cruise,and the values of each attribute are obtained by mining the taxi historical data information for the validation of the combined recommendation model.The results show that the passenger carrying hotspots obtained by clustering within the cruising range are all areas with high passenger travel demand,and the shortest driving time to reach the cruising area is inversely proportional to the results of the combined attribute values.The probability of carrying passengers on the route,the taxi supply/demand ratio in the cruising area and the expected trip revenue are positively proportional to the combined attribute values in the cruising area,which is consistent with the actual situation and proves the reasonableness of the information entropy solution weight values.Combining the actual road network traffic conditions and regional travel demand,the best cruising area is recommended for drivers to cruise,and with the cruising area determined,real-time cruising routes are recommended for drivers,and the effectiveness of the recommendation model is verified by combining taxi historical taxi data and actual empty routes.