| In recent decades,China’s cities have accelerated development,resulting in a huge demand for travel,which makes the pressure of urban passenger transport increasing.The contradiction between residents’ travel demand and supply has become serious,which restricts the development of the city.Online car hailing,also known as online taxi booking,is the application of the Internet of things in car sharing.Since the online taxi Hailing was launched,it has constantly changed the transportation structure and travel mode of the city,and has also had a great impact on the traditional taxi industry with poor service quality,high price and low efficiency.Network car Hailing has become an important part of urban passenger transport and has undertaken an increasingly arduous task of urban transport.The irrationality of the traditional scheduling scheme directly leads to the problems of long scheduling distance of drivers and passengers waiting for cancellation of orders.Travel demand is constantly transferred in space as time changes,with real-time differences.It takes a certain time for the driver to drive to a certain area.After reaching the area,the thermal map has changed,and the driver shall change the driving direction according to the new thermal map.In order to explore the subtle changes of urban taxi demand in time,shorten the waiting time of drivers and improve the scheduling efficiency,neural network needs to be used to predict the hot spots in the shortterm future in order to solve the problem of information lag.In today’s network environment,data mining and big data analysis are developing rapidly.These problems have natural application occasions and processing advantages.The huge operation data in the taxi industry strongly drives the research of this problem.This paper preprocesses the massive taxi trajectory data in Haikou,visualizes the actual location generated by online car Hailing orders,and constructs it into a spatio-temporal sequence of passenger demand point location distribution.In order to train the neural network,it is necessary to recombine the spatio-temporal sequence into a data set format in line with supervision and learning.The trained neural network can predict the scatter diagram of the location distribution of the demand points of online car hailing in Haikou in a certain period of time in the short term,and then use the K-means clustering algorithm to divide the hot spots of taxis,so the hot spots of online car Hailing are obtained.After that,this paper establishes the optimization model of no-load online car Hailing scheduling.In order to better complete the formulation and optimization of no-load network car Hailing scheduling scheme,this paper uses NSGA-Ⅱalgorithm to solve the model.Using the data prepared in the early stage,extract the coordinates and heat of passenger hot spot areas,then extract the online car Hailing location from the original data set and visualize it,then obtain the driving route planning service through Baidu map open platform,retrieve the driving route planning scheme conforming to the conditions according to the starting and ending coordinates,and return the baidu route planning mileage of online car Hailing to each hot spot area through calling the interface.Finally,put the prepared data into the NSGA-Ⅱ program to solve the model.After 100 iterations of the program,the Pareto solution set gradually tends to be stable,and all solutions gradually converge to the same curve.In the last generation Pareto solution set,an acceptable non inferior solution is randomly selected,and finally the online car Hailing scheduling scheme in a certain period of time in Haikou City is obtained.The research results have certain reference value for scientific dispatching network car hailing and alleviating people’s travel difficulties. |
