Bus Arrival Time Prediction Based On Data Mining

Developing the public transportation and increasing the rate of public transportation can alleviate the problems like urban traffic pollution and traffic congestion. Providing bus arrival time information for travelers can greatly improve the bus service satisfaction and then Increase the rate of bus travel. Providing bus arrival time information for bus managers can provide data basis for intelligent scheduling system and optimization of bus dispatching system-The bus travel time is divided into two parts:bus dwell time at bus stations and travel time between each two stations.First, This article introduce the necessity of bus arrival time prediction, summarizes the bus arrival time prediction method at home and abroad and the advantages and disadvantages of these prediction methods are analyzed.Second, because the bus GPS data only contains vehicle longitude, latitude, speed, these data cannot be directly used for bus arrival time prediction. So this paper proposes a GPS processing method, which can be used to effectively extract the bus station in time, travel time between stations and other useful information. This method of data processing is according to the regular of the bus stops running.Third, this article analyzes the influencing factors of the bus station stop time and travel time between each two stations respectively. Then, build the bus dwell time prediction model based on the average and a other model based on KNN.And then build three different prediction model for bus travel time between two stops. this paper takes two weeks bus GPS data of Nanjing 1 line for instance analysis. The analysis results show that bus dwell time prediction based on KNN is better than the other bus dwell time prediction model,and prediction model for bus travel time between two stops based on SVM is better than other models.Finally, build the bus arrival time prediction model based on the two prediction models.