Short-term traffic flow prediction model based on dynamic space-time analysis

Abstract

The invention relates to a short-time traffic flow prediction method based on dynamic space-time analysis, and the method specifically comprises the steps: system modeling: providing a dynamic traffic flow space-time state matrix for representing the traffic state of each road section of a regional road network, and precisely and quantitatively representing the traffic state of the road section; proposing a traffic flow dynamic space-time weight matrix used for calculating the similarity between traffic flow space-time state matrixes, dynamically allocating a weight to each data element of the space-time state matrixes so as to construct a dynamic space-time weighted Euclidean distance, and optimizing a neighbor search mechanism of a K neighbor model; and providing a weighted prediction method based on a similarity ratio, and carrying out empirical research by using a real traffic flow data set of an American California highway management system PeMS. The precision performance of three prediction functions including mean value prediction, inverse distance weighted prediction and grade weighted prediction is discussed. Compared with an existing statistical theory and an artificial neural network model, the model provided by the invention is verified to have better prediction precision in short-time traffic flow prediction.

Description

1. Technical field [0001] The invention relates to the field of intelligent transportation, in particular to short-term traffic flow prediction, in particular to a short-term traffic flow prediction model aiming at the dynamic spatio-temporal analysis of road network traffic flow. 2. Background technology [0002] Accurate real-time short-term traffic flow forecasting can effectively alleviate urban traffic congestion and reduce urban air pollution, which is of great social significance. Traffic flow data has characteristics such as trend, periodicity and dynamic randomness. Among them, trend and periodicity belong to the regular characteristics of traffic flow, which are mainly manifested as trends or fluctuations according to the time law, which is the premise that traffic flow can be predicted. Dynamic randomness is generated by regional road network traffic influencing factors events (such as signal lights, pedestrians passing through, road accidents, traffic control, e...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the problem that existing research does not make full use of the dynamic randomness of road network traffic flow and uses a static global fixed model structure for prediction, this invention establishes an adaptive dynamic space-time K-nearest neighbor model to realize the customized construction of each road section. modulo prediction

Images