Traffic flow prediction method based on stacked auto-encoder-support vector regression

Abstract

The invention relates to a traffic flow prediction method based on stacked auto-encoder-support vector regression. A stacked auto-encoder-support vector regression combination model is built; input traffic flow data is normalized; feature learning is carried out; data features are obtained through unsupervised layer-by-layer feature training and supervised parameter fine adjustment; the obtained data features serve as new training and testing samples; prediction is carried out through a prediction layer; finally results are inversely normalized to obtain a prediction result. According to the method, the prediction effects superior to those of other existing prediction models are achieved in the aspects of the prediction precision, the convergence speed and the like; the feasibility is high; and the effect is good. The prediction method is good in prediction effect and high in fitting degree, can effectively solve the problem of the existing prediction model, prevents the generation oflocal optimization, greatly improves the prediction precision, is wide in applicability, is suitable for traffic flow prediction of not only bridges but also common road sections, and also has a relatively good effect in the aspect of short-time traffic flow prediction.

Description

technical field [0001] The invention relates to the technical field of traffic control systems of road vehicles, in particular to a traffic flow prediction method based on stack autoencoder-support vector regression. Background technique [0002] In recent years, with the continuous acceleration of the urbanization process and the rapid growth of the number of motor vehicles, the congestion and congestion of road traffic and bridge traffic have attracted increasing attention. Accurate and real-time traffic flow information is the main basis for traffic management departments to adopt reasonable traffic planning and guidance measures, and is also the core research issue in Intelligent Transportation Systems (ITS). [0003] At present, the more common traffic flow forecasting models are mainly divided into three categories: one is more traditional forecasting models based on mathematical and physical methods, including time series models, Kalman filter models, exponential smoo...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problems existing in the prior art and overcome the shortcomings of the existing traffic flow prediction model in the prior art such as low prediction accuracy, slow convergence speed, and easy to fall into local optimization, the present invention provides an optimized stack-based autoencoder -Support vector regression traffic flow prediction method, a stacked autoencoder-support vector regression combination model is proposed, with a deep structure model as the learning layer, support vector regression as the prediction layer, and then proceed by extracting the effective data features of the input data analysis forecast

Images