Traffic speed prediction techniques in urban environments.

Area of Science:

• Transportation Engineering
• Traffic Flow Analysis
• Machine Learning Applications

Background:

• Predicting mean free-flow speed (FFS) is crucial for urban arterial traffic management.
• Existing methods may not fully capture complex interactions between traffic, geometric, and pavement variables.
• Advanced modeling techniques are needed to improve FFS prediction accuracy.

Purpose of the Study:

• To develop and compare Multiple Linear Regression (MLR) and machine learning (ML) models for predicting mean FFS.
• To evaluate the performance of Artificial Neural Network (ANN), Support Vector Machine (SVM), and Random Forest (RF) models.
• To identify the most influential variables affecting FFS on urban arterials.

Main Methods:

• Developed MLR and ML models (ANN, SVM, RF) using geometric, traffic, and pavement condition data from 11 urban arterials.
• Utilized variables such as speed, volumes, pedestrian data, lane characteristics, access points, road grade, and International Roughness Index (IRI).
• Validated models using data from two independent urban roads.

Main Results:

• Machine learning algorithms significantly outperformed traditional MLR models in predicting mean FFS.
• ML models demonstrated adaptability to traffic flow variations influenced by external conditions.
• The study identified key predictive factors for FFS, including traffic volumes and road geometry.

Conclusions:

• ML models offer a robust and accurate approach for predicting mean FFS on urban arterials.
• These models can assist in traffic management and planning, especially when direct data collection is challenging.
• The findings provide valuable insights for developing reliable traffic prediction systems.