Entropy-Weight-Method-Based Integrated Models for Short-Term Intersection Traffic Flow Prediction.

Area of Science:

• Traffic Engineering
• Data Science
• Artificial Intelligence

Background:

• Integrating prediction models is crucial for enhancing accuracy in complex systems.
• Entropy Weight Methods (EWMs) offer diverse approaches to model integration.
• Previous studies have utilized various EWMs, but their comparative performance in traffic prediction remains underexplored.

Purpose of the Study:

• To evaluate and compare the performance of three distinct Entropy Weight Methods (EWM-A, EWM-B, EWM-C) for integrating short-term traffic flow prediction models.
• To identify the most effective EWM for improving traffic flow prediction accuracy at signalized intersections.
• To analyze the strengths and weaknesses of each EWM in the context of traffic data integration.

Main Methods:

• Developed individual k-nearest neighbors (KNN) and Elman neural network models for short-term traffic flow prediction.
• Integrated the KNN and Elman models using three different Entropy Weight Methods: EWM-A, EWM-B, and EWM-C.
• Compared the predictive accuracy of the three integrated EWM models against the individual KNN and Elman models.

Main Results:

• The EWM-C integrated model consistently provided the most accurate traffic flow predictions across most evaluation days.
• The study identified specific advantages of the EWM-C method in traffic flow prediction.
• Limitations and potential issues associated with the EWM-A and EWM-B methods were highlighted.

Conclusions:

• EWM-C is a superior method for integrating short-term traffic flow prediction models at signalized intersections.
• The findings provide valuable insights for selecting appropriate EWMs in transportation modeling.
• Further research can explore refinements of EWM-C or investigate its application in other traffic prediction scenarios.