Data fusion-based traffic prediction and software decision support for recreational suburban roads

Afandizadeh, Shahriar; Abdolahi, Saeid; Mirzahossein, Hamid · 2026 · Crossref

DOI: 10.1371/journal.pone.0343224

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the challenge of predicting traffic flow on mountainous suburban roads, specifically focusing on Kandovan Road, a critical route connecting Tehran and Chalus. The research is motivated by the limitations of prior studies, which predominantly focused on urban networks and often analyzed influencing factors in isolation. Suburban mountain roads present unique difficulties due to narrow structures, hazardous segments, and highly variable conditions driven by weather, calendar events, and road-specific characteristics. To bridge this gap, the authors propose a hybrid learning framework that integrates heterogeneous data sources to model nonlinear traffic behavior and develop a software-based decision support system for real-time traffic management. The methodology involves training and evaluating eight machine learning and deep learning models: Deep LSTM, Random Forest Regressor, XGBRegressor, Transformer, ST-ResNet, Conv-LSTM, Bidirectional LSTM, and LSTM-GAN. The models were trained using a comprehensive dataset spanning from 2017 to 2023, which fused traffic data, weather information, calendar events, and road features. Performance was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Square Error (MSE), and Mean Absolute Percentage Error (MAPE). The study employed systematic hyperparameter tuning, including Grid Search and Random Search, to optimize each algorithm’s configuration. Additionally, the researchers developed a native Decision Support System (DSS) software to operationalize the best-performing models for practical use by traffic authorities. The results indicate that the Random Forest Regressor achieved the highest prediction accuracy among the evaluated models, attaining an R² score of 0.88 and demonstrating low average error metrics. This performance highlights the model’s strong capability to capture the non-linear and dynamic patterns inherent in mountainous traffic data. The study confirms that integrating diverse data sources—particularly the simultaneous consideration of weather, calendar events, and traffic behaviors—significantly enhances prediction performance compared to models relying on single data types. The findings validate the effectiveness of ensemble learning methods in handling the volatility and complexity of suburban road networks. The significance of this work lies in its contribution to sustainable and intelligent transportation planning in challenging terrains. By demonstrating that data fusion improves forecasting accuracy, the study supports more efficient traffic management and enhanced road safety. Furthermore, the development of a dedicated software application provides an operational tool for traffic authorities to visualize real-time predictions and implement intelligent action plans. This approach addresses the specific needs of recreational suburban roads, offering a scalable solution for reducing congestion, minimizing fuel consumption, and mitigating the risks associated with unauthorized road usage in hazardous environments.

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.

• telematics crash prediction