Design Of Bus Arrival Time Prediction Algorithm Based On Macnine Learning

Vigorously construction of public transportation helps to reduce the pressure on urban traffic and environment, intelligent bus dispatch and monitoring system can significantly improve the public transportation service and management level which lead to better service for the public and attract more people to choose public transport. Bus arrival time prediction system helps to improve the service of the intelligent transport system. This paper studies how to build the bus arrival time prediction model and algorithm.Firstly this paper expounds the composition of the intelligent bus dispatch and monitoring system and discusses what the position is of bus arrival time prediction in the intelligent bus dispatch and monitoring system, and at the same time introduces current research achievement of bus arrival time prediction algorithms. Then analyzes the weather and human factors that have influences on bus running with real bus running data. Analyzes the law of bus running time in different time of a day with real data and clustering algorithms. Analyzes the law of historical bus running time and discusses how to choose the historical data. Give out an bus average speed prediction model for residual road to station based on the law of bus running speed.After sufficient analysation of factors and laws that have influences on bus running, the paper confirms the factors and reference data that should be the input of the prediction algorithm. According to the status of bus running on road, present a prediction model that is composed of three sub prediction model, and design the input data structure of each model. Confirms the model presented by this paper can work effectively by training and testing the model with real running data, and the predicting precision is higher than the models using mean value of historical and pre-buses running data. And realizes the prediction model with the design scheme.