Study Estimating hourly traffic flow using Artificial Neural Network: A M25 motorway case

Turki, Ahmed Ibrahim; Hasson, Saad Talib · 2023 · Crossref

DOI: 10.54153/sjpas.2023.v5i1.448

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 estimating hourly traffic flow on highways, a critical requirement for transportation agencies to calculate network-wide performance metrics such as congestion and user delay costs. Accurate flow estimation is difficult because vehicle counts are typically available only at sparse, fixed sensor locations, and existing hourly profiles often fail to account for variations caused by weather, incidents, or daily demand fluctuations. To overcome these limitations, the authors propose a hybrid strategy that combines machine learning with technical analysis indicators, aiming to improve short-term traffic flow prediction accuracy. The methodology utilizes daily traffic data from the Motorway Incident Detection and Automatic Signalling (MIDAS) system on the M25 motorway in the United Kingdom, specifically between Junctions 13 and 14. The raw data, recorded every 15 minutes, was cleaned to handle noise and missing values before being used to extract six specific technical indicators: Average True Range (ATR), Simple Moving Average (SMA), Exponential Moving Average (EMA), Relative Strength Index (RSI), Rate of Change (ROC), and Momentum (MOM). These indicators served as input features for a fully-connected, multi-layer Artificial Neural Network (ANN). The ANN architecture consisted of an input layer with 40 neurons, two hidden layers with 60 neurons each, and a single output neuron. The model employed Exponential Linear Unit (ELU) activation functions to accelerate learning and dropout regularization to prevent overfitting. Training was conducted using the Adam optimization algorithm with a learning rate of 0.001 and a batch size of 32 over 1,000 epochs. The model was trained on four days of data and tested on a subsequent day, with performance evaluated using Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). The results demonstrate that the proposed ANN model significantly outperformed other machine learning techniques, including Gradient Boosting, Random Forest, and Linear Regression. The ANN achieved an RMSE of 1.3 and an MAE of 1.09, whereas Random Forest and Gradient Boosting yielded substantially higher error rates (RMSE of 95.9 and 54.1, respectively). Although Linear Regression showed a low median RMSE of 1.01, the ANN was favored for its ability to detect pattern changes quickly and its overall robustness. A sensitivity analysis revealed that the Simple Moving Average (SMA) was the most critical input feature; its exclusion caused a drastic increase in error (RMSE rising to 74.45). Conversely, the Relative Strength Index (RSI) was the least impactful, as its removal resulted in minimal performance degradation. The study concludes that integrating technical analysis indicators with machine learning models enhances predictive power for traffic flow estimation. The proposed hybrid approach offers a simple yet effective method for generating accurate hourly flow estimates from limited sensor data. The authors suggest that future iterations of this model could incorporate additional parameters, such as traffic density, speed, and weather conditions, to further refine forecast precision. This work provides a viable solution for transportation agencies seeking to improve the accuracy of their performance measurements and traffic management 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