The Bus Dispatching Method Based On Big Data And Hybrid Heuristic Algorithm

Since the Intelligent Transportation System(ITS)being used in our country from the beginning of this century,the traffic management department and operational companies have been accumulated massive amounts of data.These data include the movement track of the bus,passengers data,bus rapid transit station data,the movement track of the taxi,public bicycle borrowed records,etc.The value of the data is huge,but we have not been used it yet.Even though part of them have been discarded because of the costs to storage is too high,and also because there really too much data generated.In this paper,we will introduce how to processing and analysis these data,and to improve the operation efficiency and service quality by using the data mining method,adaptive particle swarm optimization(APSO),BP neural network and genetic algorithm.In this paper,we describe the source of the data in the bus system and introduce the format of the smart card data and bus GPS history records.Before data mining,it is necessary to preprocess the abnormal data in the data set.These abnormal data include invalid data,redundant data,and the absence of the up or down station of the smart card data,the inaccurate data of the GPS records.After the processing,it's need to normalization the numerical data,and convert the category data to a numerical value.Then cluster the smart card data in different dates used by the hierarchical clustering analysis with the Ward method.Compared with the previous method,this method is more concerned about the overall rule of the traffic flow in a day,which is more accordance with the actual situation.Finally,the APSO-BP neural network is used to predict the traffic flow,with the parameters of the time period,the weather,the holidays and the date.In the establishment of forecasting model,this paper puts forward that the weather data is divided into three grades according to the influence of the public transportation.For the GPS data,the relevant operational characteristics are analyzed,and then the APSO-BP method is used to predict the arrival time of the vehicle.There are two very important activities in public transportation scheduling,Static scheduling and dynamic scheduling.Solutions are puts forward in these two aspects in this paper.Based on the passenger flow data that is forecasted,using the improved simulated annealing and genetic algorithm to solve the bus scheduling problem.Inthis paper,the individual fitness is used to improve genetic algorithm.It will improve the shortcomings of the genetic algorithm,by increasing the diversity of the individuals in early stage,and speed up the convergence in later period.The mathematical model of static scheduling is established with the objection of minimum the passengers waiting cost and the maximum the income of public transportation operation organization.In the aspect of dynamic scheduling,the exceptional situation can be detected by the forecasted arrival time and the real-time location information of the bus.It can also conduct the bus priority scheduling based on the number of the onboard passengers.Combining the results of the static schedule,the collaborative schedule can be conducted with other transportation or routes.