Short-time traffic flow forecast method based on grey wavelet neural network

Abstract

The invention relates to a short-time traffic flow forecast method based on a grey wavelet neural network. The method comprises the steps that according to a time sequence of initial traffic flow, a single-factor grey forecast model is applied to obtain an initial forecast time sequence; a pre-forecast time period and a post-forecast time period are set; the wavelet neural network is applied to pre-forecast the pre-forecast time period to obtain a pre-forecast value and a pre-forecast time sequence; according to the pre-forecast time sequence, a maximum Lyapunov index method is applied to forecast the post-forecast time period to obtain a post-forecast value; aggregation is carried out on the pre-forecast value and the post-forecast value to obtain a forecast value of the traffic flow. Thesingle-factor grey forecast model is adopted, the stochastic fluctuation of traffic flow time sequence data is greatly weakened, the error fluctuation of the forecast value is small, post-forecastingis carried out by adopting the maximum Lyapunov index method, the error fluctuation caused by chaotic motion in the later period of pre-forecast data of the wavelet neural network is reduced, and thestability and fitting degree of the forecast value of the short-time traffic flow are improved.

Description

technical field [0001] The invention relates to a traffic data management technology, in particular to a short-term traffic flow prediction method based on gray wavelet neural network. Background technique [0002] Urban road traffic flow prediction is an important part of Intelligent Transportation System (ITS). Timely, effective and accurate traffic flow prediction plays an important role in intelligent traffic induction and road traffic congestion relief. [0003] Short-term traffic flow has obvious volatility and randomness, so it is necessary to fully consider its random fluctuation and weak regularity when predicting short-term traffic flow. For the short-term prediction of traffic flow, the researches of relevant scholars at home and abroad are mainly divided into two types: prediction based on mathematical models and prediction based on non-mathematical models. Among them, the prediction based on mathematical models mainly refers to the improved calculation and comb...

AI Technical Summary

Problems solved by technology

Among them, the neural network can describe the internal fluctuations of any complex system, and can perform simulation operations based on historical data. However, when the sample data and computing units are relatively limited, and the original data has large random fluctuations, the stability of the prediction results is poor. , and it is easy to fall into a local optimum

Images