Abstract:Traffic flow is chaotic due to nonstationary realistic factors, and revealing the internal nonlinear dynamics of chaotic data and making high-accuracy predictions is the key to traffic control and inducement. Given that high-quality phase space reconstruction is the foundation of predictive modeling. Firstly, an improved C-C method based on the fused norm search domain is proposed to address the issue that the C-C method in the phase space reconstruction algorithm does not meet the Euclidean metric accuracy and reduces the reconstruction quality when the infinite norm metric is used. Secondly, to address the problem of insufficient learning ability of traditional convolutional combinatorial modeling for complex phase space laws of chaotic traffic flow, the high-dimensional phase space features are extracted using the layer-by-layer pretraining mechanism of convolutional deep belief networks (CDBNs), and the temporal features are extracted by combining with long short-term memory (LSTM). Finally, an improved probabilistic dynamic reproduction-based genetic algorithm (PDRGA) is proposed to address the problem of the hybrid model falling into a local optimum when learning the phase space law. Experiments are conducted in three aspects: phase space reconstruction quality analysis, comparison of optimization algorithm convergence, and prediction model performance comparison. The experimentation with two data sets demonstrates that the improved C-C method combines the advantages of the high accuracy metric of the L2 norm with the low operational complexity of the infinite norm, achieving a balance between reconstruction quality and algorithm efficiency. The proposed PDRGA optimization algorithm is a lightweight improvement of the traditional genetic algorithm (GA) and solves the problem that the model tends to fall into a local optimum by optimizing the initial weights of CDBN. Meanwhile, the five error evaluation indexes of the proposed PDRGA-CDBN-LSTM hybrid model are lower than those of the baseline model, providing a new modeling idea for chaotic traffic flow prediction.

Self-similar traffic prediction based on Lyapunov exponent

Prediction of Self-Similar Traffic and Its Application in Network Bandwidth Allocation

The On-Line Prediction of Self-Similar Traffic Based on Chaos Theory

A Combined Algorithm for Real-time On-line Rapid Identification of the Chaos in Traffic Flow

Modeling and Simulation of Self-Similar Traffic Based on FBM Model

Identification of Chaos in the Traffic Flow Based on the Improved Largest Lyapunov Exponents Algorithm

The Short-Term Predictability Of Traffic Flow

The Computing of Maximum Lyapunov Exponent in Traffic Flow Applying the Improved Small-data Method

Chaos Characteristics of Expressway Traffic Flow

Traffic Flow Predicting Of Chaos Time Series Using Support Vector Learning Mechanism For Fuzzy Rule-Based Modeling

Chaotic Characteristic Analysis for Traffic Flow Series and Dfpsovf Prediction Model

Study on transformation process of chaos in simulation traffic flow based on distance headway

Simulation Study on Influence Factors of Transition of Chaos in Traffic Flow Based on Car-following Model

Relationship between Distance Headway and Chaos in Traffic Flow on Expressway

The Study of Short-Term Traffic Flow Forecasting Based on Theory of Chaos

A method of on-line fast identification for the chaos in traffic flow

Joint Network Traffic Forecast with ARIMA Models and Chaotic Models Based on Wavelet Analysis

An Improved Phase Space Reconstruction Method-Based Hybrid Model for Chaotic Traffic Flow Prediction

Network Traffic Prediction Algorithm and Its Practical Application in Real Network

Simulation study of the transition between chaos and order motion in traffic flow

Modeling Self-Similar Traffic for Network Simulation