Modeling Based On Multi-source Data Of Bus Arrival Time

With the acceleration of the modernization process,establishing an intelligent public transportation system has become a meaningful way to solve the urban traffic problems.As one of the most concerned problems of passengers and traffic departments,the accuracy of bus arrival time prediction is the key to realize an intelligent public transportation system.However,many cities in China（such as Beijing）have numerous bus routes and complex traffic conditions.Under such circumstances,it has become a pressing issue to improve the accuracy of bus arrival time prediction.In addition,due to abnormal stops during the bus journey,bus accidents occur frequently,and passengers’ safety is difficult to be guaranteed.Therefore,managing the bus’ s abnormal stop behavior intelligently has also attracted people’s attention.This paper studies the above problems of buses,and the specific research contents and innovative results are as follows:For bus arrival time prediction,most studies use linear interpolation to increase the auxiliary points of the model in practical application to monitor the changes of bus traffic state at a higher frequency.However,additional interpolations cause the data dimensions to increase rapidly,and not all interpolations are of equal importance.Therefore,we propose a time series prediction model based on Deep Learning to select interpolation points.The model selects more characteristic interpolation points by interacting with the Reinforcement Learning network and arrival time prediction network.It can effectively improve the efficiency and accuracy of bus arrival time prediction.Experimental results show that our model has higher accuracy compared with other bus arrival prediction algorithms.For the problem of bus abnormal stops,we propose a bus abnormal stop judgment model based on low-rank matrix decomposition.Firstly,we propose a method to calculate the stopping points to obtain all the possible stopping points on the bus journey,and construct the stopping point matrix.Then,we establish a bus abnormal stop judgment model.Through analyzing low rank property of abnormal stop points,the concrete steps of the algorithm to restore abnormal stop points are given.The experimental results show that the bus abnormal stop judgment model can effectively capture the abnormal stop position,and effectively improve the accuracy of the bus abnormal stop point judgment.