On-line Prediction Of Bus Arrival Time Based On Support Tensor And Dynamic Hybrid Algorithm

With the rapid development of social economy,city residents living standards continue to improve,the amount of vehicle and road traffic volume is increasing rapidly,is to bring the problems of traffic congestion,environmental pollution,energy waste,and accelerate the development of public traffic is an effective way to solve these problems.The key to accelerating the development of public transport is to provide better service for travelers,bus arrival time prediction to make travel passengers can arrange travel plans,is an effective way to improve the quality of public transport services.In this paper,the research status of domestic and foreign arrival time prediction technology is analyzed and compared,and the development of bus information collection technology and some related theoretical knowledge are introduced.And then use the existing historical data bus,analyzes the influencing factors of bus arrival;using the sample data combining with clustering analysis algorithm,analyzes the bus travel time of mining the hidden data in the sample prior information;analysis of the law of historical data and real-time data sample,discusses the selection of input features,and draw some methods and strategies for feature selection.After analyzing the factors which influence the arrival time of the bus,a new bus arrival time prediction algorithm based on support vector machine and dynamic hybrid algorithm is proposed.The algorithm is a prediction method to combine static and dynamic online prediction,static prediction algorithm is a prediction algorithm for modeling support tensor machine based on the characteristics of the input were time,whether working days,weather,road length and speed,combining with clustering analysis to time information,make full use of the characteristics of the data structure without losing support tensor machine when modeling sample,and improve the accuracy of the prediction results.Dynamic prediction according to the running state of the bus,we constructed three sub models to predict,then the real-time data set with dynamic hybrid algorithm,the reference time and the use of static,realize the prediction of bus arrival time online.Finally,the model is verified by the sample data of K1 bus line in Ganzhou city.Results show that the prediction research and historical average and the average bus arrival time online support tensor machine and dynamic mixed algorithm based on comparing the forecast accuracy is significantly improved and the obvious enhancement of stability prediction.