Traffic flow prediction method based on improved time-space association KNN (K-Nearest Neighbor) algorithm

Abstract

The invention provides a short-time traffic flow prediction method based on an improved time-space association KNN (K-Nearest Neighbor) algorithm. The core concept of the invention is that firstly, astate vector capable of reflecting a change trend of traffic flow is constructed; further, K state vectors which are the nearest neighbor to a current time period of a target detector are searched from two dimensions of a time domain and a space domain of historical state vectors; weight allocation of the K state vectors is analyzed; and a flow value at a next moment is predicted. According to theinvention, a short-time flow sequence of an intersection can be excellently predicted, the technical support is provided for improving intelligence and scientificity of flow prediction and promotingoperation efficiency of traffic flow of the intersection, and the short-time traffic flow prediction method belongs to the field of the traffic control study.

Description

technical field [0001] The present invention proposes a short-term traffic flow prediction method based on the improved spatio-temporal correlation K-Nearest Neighbor (KNN) algorithm, which predicts the traffic flow passing through the target detector in the traffic road network in the next period, and supports traffic management and signal Control strategy formulation and program optimization belong to the field of intelligent transportation research. Background technique [0002] As a data-driven method, the KNN algorithm does not require complex mathematical models or prior knowledge. Due to the flexibility of the method, it has been widely used in nonparametric methods. The basic idea of ​​the traditional KNN algorithm is to find the K sequences closest to the traffic mode in the current period of the target detector from all historical traffic flow sequences, and further obtain the traffic values ​​of the K sequences at the next moment and compare them according to the...

AI Technical Summary

Problems solved by technology

However, most of the existing methods only consider the time domain, and only use the traffic flow data of the current period and the historical period of the target detector to predict the traffic flow of the next period, but in the same period, the target road section will interact with its upstream and downstream road sections, so It is a feasible method to improve the KNN algorithm by increasing the spatial correlation of traffic flow

Images