WEKA-based machine learning for traffic congestion prediction in Amman City

Arabiat, Areen; Hassan, Mohammad; Almomani, Omar · 2024 · Crossref

DOI: 10.11591/ijai.v13.i4.pp4422-4434

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 Amman City, specifically at the 8th Circle intersection, which connects four main streets: Westbound, Northbound, Eastbound, and Southbound. Motivated by the societal and economic costs of congestion, including wasted time, pollution, and fuel consumption, the research aims to predict traffic status using machine learning (ML) techniques. The authors seek to identify the most effective ML classifier for this specific urban environment, leveraging historical traffic data to improve traffic management and reduce inefficiencies. The methodology utilizes hourly traffic data collected from the Greater Amman Municipality for the entire year of 2019. The dataset includes attributes such as traffic volume, density, speed, occupancy, width, and distance for each lane and approach. After cleaning and preprocessing the data to remove duplicates and structural errors, the authors employed the Waikato Environment for Knowledge Analysis (WEKA) tool to evaluate six ML classifiers: Logistic Regression (LR), K-Nearest Neighbor (KNN), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP). The experimental design involved splitting the data into 70% for training and 30% for testing, supplemented by 10-fold cross-validation. Performance was assessed using accuracy, precision, sensitivity, and F-measure metrics derived from confusion matrices. The results demonstrate that the Support Vector Machine (SVM) classifier consistently outperformed the other algorithms across all four street approaches. SVM achieved the highest accuracy rates: 99.4% for Westbound Street, 99.7% for Northbound Street, 99.6% for Eastbound Street, and 99.1% for Southbound Street. In contrast, the Multi-Layer Perceptron (MLP) classifier yielded the lowest accuracy, ranging from 94.5% to 96.3%. SVM also recorded the highest scores for precision, sensitivity, and F-measure across the experiments. For instance, on the Northbound Street, SVM achieved a precision of 99.8% and an F-measure of 99.7%, significantly surpassing the performance of DT, RF, KNN, LR, and MLP. The significance of this study lies in its demonstration that SVM is the superior classifier for predicting traffic congestion in this specific urban context, achieving accuracy rates near 99.7%. These results are notably higher than those reported in previous studies, which typically ranged between 84% and 91% using models like LSTM or linear regression. The findings suggest that SVM-based models can provide highly reliable predictions for traffic management systems, potentially aiding in reducing congestion-related costs and environmental impacts. The study validates the effectiveness of using WEKA for data mining in smart city applications and highlights the importance of selecting appropriate classifiers for specific traffic datasets.

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