Short-term traffic flow prediction research based on ICEEMDAN-MPE-PSO-DELM model.

Area of Science:

• Traffic Engineering
• Artificial Intelligence
• Time Series Analysis

Background:

• Accurate short-term traffic flow prediction is crucial for intelligent transportation systems.
• Existing models often struggle with the complex, non-linear dynamics of traffic flow at intersections.

Purpose of the Study:

• To propose a novel hybrid model for improving short-term traffic flow prediction accuracy at intersections.
• To leverage data decomposition and ensemble learning for enhanced prediction performance.

Main Methods:

• Empirical Mode Decomposition (EMD) variants like ICEEMDAN for time series decomposition.
• Multi-scale Permutation Entropy (MPE) with Particle Swarm Optimization (PSO) to assess randomness of components.
• Hybrid prediction using Deep Extreme Learning Machine (DELM) and ARIMA models based on component randomness.

Main Results:

• The proposed ICEEMDAN-MPE-PSO-DELM-ARIMA model demonstrated superior performance.
• Achieved the smallest prediction errors compared to other benchmark models.
• Exhibited the best fitting effect with actual traffic flow values.

Conclusions:

• The novel hybrid model effectively improves short-term traffic flow prediction accuracy.
• The combination of decomposition, randomness assessment, and ensemble prediction is highly effective.
• The model offers a promising solution for intelligent traffic management.