From Numerical Simulations to Machine Learning Algorithms: Traffic State Prediction of Mixed Car-Bike Flow
Research Poster Engineering 2025 Graduate ExhibitionPresentation by Faeze Fathijam
Exhibition Number 26
Abstract
This research aims to understand the use of machine learning models when predicting traffic conditions in mixed traffic flow. By utilizing a dataset created through a numerical simulation of traffic flow theory, various Machine Learning models, like Random Forest (RF) and Neural Networks (NN) are trained and assessed to understand their ability to accurately predict capacity and traffic delay. The models are assessed using various metrics, with a focus on R² scores to determine their predictive accuracy. Our findings reveal that machine learning methods, especially Random Forest, can accurately predict traffic flow and delays in traffic environments. The results also suggest that these machine learning models can perform well when predicting traffic conditions not present in the initial dataset. In other words, these models can extrapolate well, especially if some congested conditions are present in the initial training dataset. This study contributes to enhancing the efficiency of traffic control systems and underscores the probable application of data driven strategies in comprehending traffic behaviors.
Importance
Traffic congestion, especially in mixed traffic conditions with cars and bicycles, poses a significant challenge for urban mobility. This study integrates machine learning with traffic flow theory to predict traffic conditions and delays, using models such as Random Forest and Neural Networks. By analyzing large-scale simulated traffic data, the research identifies key factors influencing congestion and evaluates model performance in both seen and unseen conditions. The findings offer insights for traffic engineers and policymakers, improving data-driven traffic management and enhancing urban mobility. This research contributes to the development of efficient, scalable, and adaptive transportation solutions that reduce congestion and improve travel time reliability.