Real-time bus passenger flow prediction method based on neighbor regression

Abstract

The invention discloses a real-time bus passenger flow prediction method based on neighbor regression, and utilizes the passenger flow history and the real-time data of a bus route and a bus stop to predict the bus passenger flow in a future time period. Firstly, a database which owns a great quantity of historical bus passenger flow data is established, the data is subjected to data preprocessingto guarantee the data quality of a sample, and a neighbor regression algorithm is used for setting the distance measurement way, the neighbor number K and the prediction formula of a model. A passenger flow number in a future time period is predicted according to a current real-time passenger flow number, prediction accuracy is high, the method is simple in operability and universality, the problem that the bus passenger flow is nonlinear and complex can be solved, a powerful basis is provided for bus dispatching and reasonable scheduling, and meanwhile, the travel satisfaction degree and comfort level of passengers can be improved.

Description

technical field [0001] The invention relates to the technical field of bus passenger flow forecasting, in particular to a real-time bus passenger flow forecasting method based on neighbor regression. Background technique [0002] In the field of transportation, big data has always been regarded as a technological weapon to relieve traffic pressure. With the development of mobile phone networks, GPS / Beidou car navigation, Internet of Vehicles, and Internet of Things for transportation, information on people, vehicles, and roads in transportation elements can be collected in real time, and the sources of urban traffic big data are increasingly abundant. Bus passenger flow is an important data indicator that can reflect people's travel patterns and bus load. Accurate and effective bus passenger flow forecasting not only provides a strong basis for reasonable bus scheduling. [0003] The salient feature of bus passenger flow data is its high degree of nonlinearity and uncertain...

AI Technical Summary

Problems solved by technology

Traditional statistical methods such as time series models rely on the quality of historical data and cannot fully consider the uncertainty of passenger flow data. Such prediction methods have low accuracy and low reliability

Machine learning prediction methods such as neural networks and support vector machines can improve prediction accuracy, but the model complexity is too high, the training time is too long, and parameter selection is difficult

Images