Comparative Application of Radial Basis Function and Multilayer Perceptron Neural Networks to Predict Traffic Noise Pollution in Tehran Roads

Mansourkhaki, Ali; Berangi, Mohammadjavad; Haghiri, Majid · 2018 · Crossref

DOI: 10.12911/22998993/79411

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 noise pollution, a significant environmental hazard that threatens human health but lacks effective control measures due to complex urban variables. The authors aim to evaluate and compare the performance of two artificial neural network architectures—Multilayer Perceptron (MLP) and Radial Basis Function (RBF)—in estimating the equivalent continuous sound level ($L_{Aeq}$) on roads in Tehran, Iran. The motivation stems from the need for accurate, intelligent prediction models that can handle the non-linear and heterogeneous nature of urban noise data, potentially offering superior alternatives to traditional regression models. The research methodology involved collecting field data from 34 locations in the west and northwest of Tehran, resulting in 51 samples. Measurements were taken between 7 a.m. and 8 p.m. on working days using a sound level meter positioned 1.2 meters above the road surface and 2 meters from the carriageway edge. The input variables for the models included traffic volume, average vehicle speed, and the percentage of heavy vehicles. The dataset was randomly divided into training (80%), validation (10%), and testing (10%) subsets. The MLP model was trained using the Levenberg-Marquardt algorithm, while the RBF model utilized Gaussian activation functions with parameters optimized through trial and error. Both models were implemented in MATLAB R2014b, and their performance was assessed using Mean Squared Error (MSE) and the coefficient of determination ($R^2$). The results demonstrated that both neural networks provided accurate predictions, but the MLP model outperformed the RBF model. The optimal MLP architecture, featuring seven neurons in the hidden layer, achieved an MSE of 0.6292 and an $R^2$ of 0.947, with prediction errors ranging between -1.99 and +1.92 dB. In contrast, the RBF model, which required 41 neurons in the hidden layer to minimize error, yielded a higher MSE of 1.786 and a lower $R^2$ of 0.8626, with errors ranging from -5.62 to +1.19 dB. The MLP model showed a stronger correlation (0.973) compared to the RBF model (0.928), indicating a better fit between predicted and measured noise levels. The significance of this study lies in confirming the efficacy of artificial neural networks for traffic noise prediction and establishing the superiority of MLP over RBF for this specific application. The findings suggest that MLP networks offer higher accuracy and generalization capability with fewer hidden neurons, making them a more efficient tool for environmental noise assessment. This contributes to the broader field of ecological engineering by providing a robust method for monitoring and managing urban noise pollution, aiding in the development of clearer standards and control strategies.

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