RegTraffic: A Regression based Traffic Simulator for Spatiotemporal Traffic Modeling, Simulation and Visualization

Mostafi, Sifatul; Alghamdi, Taghreed; Elgazzar, Khalid · 2022 · Crossref

DOI: 10.1109/ijcnn55064.2022.9892185

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper introduces RegTraffic, a novel regression-based traffic simulator designed for spatiotemporal modeling, simulation, and visualization of road traffic. The research addresses the limitations of existing microscopic and macroscopic simulators (such as SUMO, VISSIM, and Aimsun), which often lack the ability to simulate congestion based on the intercorrelation of neighboring road links using small amounts of real-time data, and frequently lack interactive web-based visualization. RegTraffic aims to fill this gap by predicting the congestion of a specific road link based on the traffic conditions of connected inbound links and temporal features, providing a user-friendly interface for dynamic analysis. The methodology employs a Bayesian linear regression approach to model the dependent spatial feature (an outbound road link) as a function of independent spatial features (inbound road links) and temporal features. Temporal features, such as time of day and peak hours, are extracted via exploratory data analysis and encoded categorically. The model incorporates event integration, allowing specific events to override standard spatial feature values within defined time constraints. The system’s processing pipeline extracts time-series congestion index data (average speed in km/h) from Google Maps every 15 minutes. Users can define routes, select dependent and independent features, and input new observations for independent variables to predict outcomes for the dependent link. The study evaluates RegTraffic using data from four connected road links in Oshawa, Ontario, Canada, collected over one week in March 2020. Through exploratory data analysis, temporal features like "Peakhour" (speed below 11.75 km/h) and "AM" were identified. The Bayesian linear regression model was compared against Multiple Linear Regression, Elastic Net Regression, and a baseline mean model. Performance evaluation using Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) demonstrated that RegTraffic achieved an MAE of 1.3123 km/h and an RMSE of 1.71962 km/h. These results outperformed the other regression models and significantly outperformed the baseline, which had an MAE of 3.75357 km/h and an RMSE of 5.09258 km/h. The visualization component successfully displayed predicted congestion on interactive geographical maps. The significance of this work lies in providing a lightweight, regression-based alternative to complex agent-based simulators, particularly useful for scenarios with limited real-time data. By integrating Bayesian inference, RegTraffic offers not just point estimates but probability distributions that quantify uncertainty in traffic predictions. The inclusion of an interactive web interface and geographical visualization makes the tool accessible for planning and management purposes, allowing users to dynamically simulate traffic flows and understand the impact of neighboring road conditions on specific segments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• mental model of traffic

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model