Prediction of Traffic Flow in Vehicular Ad-hoc Networks using Optimized Based-Neural Network

Turki, Ahmed Ibrahim · 2024 · Crossref

DOI: 10.54153/sjpas.2024.v6i2.758

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 accurately predicting traffic flow in Vehicular Ad-hoc Networks (VANETs), a critical component of Intelligent Transportation Systems for managing congestion, reducing pollution, and ensuring public safety. The research is motivated by the complex, non-linear, and unpredictable nature of traffic patterns, which are often influenced by unforeseen variables such as accidents or road closures. Existing methods, including conventional statistical models and deep learning techniques like Recurrent Neural Networks, often struggle with computational efficiency or fail to fully capture dynamic spatial-temporal dependencies. To overcome these limitations, the authors propose a novel prediction strategy that enhances the convergence speed and accuracy of feedforward neural networks by replacing the standard Gradient Descent algorithm with the Quasi-Newton optimization method. The experimental design utilizes historical traffic data from three distinct datasets: Highway England (HE), the Performance Measurement System (PeMS) in California, and Maryland 511, covering the period of January 2023. The methodology involves several preprocessing steps, including data cleaning, smoothing, and normalization. Crucially, the model incorporates four technical indicators—Average True Range (ATR), Exponential Moving Average (EMA), Relative Strength Index (RSI), and Rate of Change (ROC)—as input features to capture volatility and trends. The core predictive model is an Optimized Multilayer Perceptron (O-MLP) with one hidden layer containing three neurons and a hyperbolic tangent activation function. This model is trained using the Backpropagation algorithm optimized by the Quasi-Newton method, specifically leveraging the Broyden-Fletcher-Goldfarb-Shanno (BFGS) update rule to approximate the Hessian matrix, thereby accelerating convergence. The performance of the O-MLP is evaluated using Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) and compared against K-Nearest Neighbors (KNN) and Naive Bayes baselines. The results demonstrate that the proposed O-MLP model significantly outperforms the comparative methods across all three datasets. For the Highway England dataset, the O-MLP achieved an MAE of 1.07 and an RMSE of 0.10, whereas KNN and Naive Bayes yielded substantially higher errors (MAE of 5.91 and 7.07, respectively). Similarly, for the PeMS and Maryland 511 datasets, the O-MLP recorded the lowest error metrics, with an RMSE of 0.07 for Maryland 511. The study reports that the O-MLP improved RMSE performance by approximately 98.58% compared to KNN and 98.88% compared to Naive Bayes for the Highway England data. These findings indicate that the integration of technical indicators with Quasi-Newton optimization effectively reduces the error factor and expedites the learning process. The significance of this research lies in its contribution to more efficient and accurate short-term traffic flow forecasting tools for transportation agencies. By demonstrating superior performance over established machine learning algorithms, the proposed model offers a robust solution for real-time traffic management and congestion mitigation. The authors conclude that while the model is effective, future work should address exogenous factors such as seasonal variations and accidents, potentially by integrating Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs) to handle broader spatial-temporal dynamics.

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.

• anticipation