Data-Driven Optimization For Public Transportation System

With the rapid growth of population and limited living space,most of the world’s major cities,such as Shenzhen,Singapore and Sydney,are increasingly dependent on the public transport system in the city.Because of the complex traffic situation in the city and the constantly changing travel needs of the residents,the taxi cannot accurately match the travel needs of the passengers,and the bus is difficult to operate in the urban road according to the planned operation schedule in advance.The random cruising of taxis in the city often results in the waste of transportation capacity and traffic jams The bus cannot run according to the planned schedule,which often results in the situation that two consecutive buses of the same route arrive at the same bus station at the same time or at a very close time("bus bunching"),thus increasing the waiting time of passengers.These issues significantly reduce the operation efficiency of the public transportation system.It is of great significance to study how to resolve these problems in taxi and bus system to improve the efficiency of urban operationIn recent years,more and more attention has been paid to the taxi ridesharing Different from other previous studies,we fully mine the mobility information of passengers and taxis,and propose the mT-Share system,which uses geographic information and mobility information to index and match passengers and taxis,and at the same time supports taxis to proactively find offline passengers for services.We use a real public dataset to evaluate the performance of the mT-Share system.The results show that the system can respond to the travel demand submitted by each passenger in milliseconds in real time.When the detour distance of each passenger increases slightly,it can serve 42%more passengers in peak time and 62%more passengers in off peak time than other methodsIn the bus system,bus bunching results in the waste of transportation capacity and inefficient operation of the system.The key step to solve this problem is to be able to accurately and timely predict when and which station the next bus bunching will occur for a specific bus line in the city.Therefore,in this study,we propose SD-seq2seq model,which predicts whether this phenomenon will occur at all stops of a bus line in the future by extracting the features from the supply side and demand side of the bus operation process as input.We use the data of the Sydney smart card data to evaluate the performance of this model.Through the experiments of multiple bus routes,we prove that the SD-seq2seq model can achieve more than 85%prediction accuracy.