An evaluation of multiple classifiers for traffic congestion prediction in Jordan

Hassan, Mohammad; Arabiat, Areen · 2024 · Crossref

DOI: 10.11591/ijeecs.v36.i1.pp461-468

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 traffic congestion in urban environments, specifically focusing on the 8th Roundabout in Amman, Jordan. Traffic congestion causes significant economic losses, environmental pollution, and reduced productivity. While infrastructure improvements are common solutions, the authors argue that leveraging Internet of Things (IoT) and machine learning (ML) technologies offers a more effective approach for predicting congestion patterns and optimizing traffic flow. The research aims to identify the most effective ML classifier for predicting traffic congestion by evaluating multiple algorithms against real-world traffic data. The methodology utilizes a dataset provided by the Greater Amman Municipality, covering traffic conditions from January 1, 2019, to December 31, 2019. The data includes variables such as traffic volume per lane, density, speed, occupancy, width, and distance, collected via sensors at four approaches (westbound, northbound, eastbound, and southbound) to the roundabout. Each approach contained approximately 8,640 records. The data was preprocessed using Excel to remove duplicates and correct structural errors, then converted to CSV format for analysis in the WEKA data mining tool. The study evaluated seven classifiers: Naïve Bayes (NB), Stochastic Gradient Descent (SGD), Fuzzy Unordered Rule Induction Algorithm (FURIA), Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Multi-Layer Perceptron (MLP). Models were trained using a 70/30 train-test split and validated with 10-fold cross-validation. Performance was assessed using accuracy, precision, sensitivity, and F-measure. The results demonstrate that the FURIA classifier significantly outperformed the other six models. FURIA achieved 100% accuracy, precision, sensitivity, and F-measure across all experimental scenarios. In comparison, the next best performer, SGD, achieved 97.91% accuracy, while NB performed the lowest with 87.54% accuracy. The study notes that these results surpass the accuracy rates reported in previous literature, which ranged from 84% to 91% for models like LSTM, Random Forest, and ConvLSTM. The authors attribute FURIA’s superior performance to its ability to generate fuzzy rules and utilize an effective rule stretching mechanism, allowing it to differentiate classes efficiently. The significance of this research lies in providing urban planners and policymakers with a highly accurate tool for traffic management. By identifying FURIA as the optimal classifier for this specific context, the study offers a pathway to reduce fuel consumption, time wastage, and emissions through better traffic control. The authors conclude that while the current model is highly effective, future studies should incorporate additional variables, such as weather conditions and driver behavior, to further enhance prediction accuracy and robustness.

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.

• traffic density