An Optimized Multi-layer LSTM Network for Real-Time Short-Term Traffic Forecasting in Urban Environments

DOI

10.2991/978-94-6239-664-7_83How to use a DOI?

Keywords

LSTM; Short-term traffic flow prediction; RNN

Abstract

To alleviate traffic congestion in the fast-urbanizing cities such as Dhaka, Bangladesh, real time prediction of short-term traffic flow should be accurately predicted. Although deep learning models like LSTM networks are great at the ability to capture the time dynamics, they become weak when exposed to non-smooth and non-stationary data of urban traffic. The proposed paper suggests an optimized multi-layer LSTM network as a solution to these problems through a segment based adaptive optimization process. Its architecture uses stacked LSTM layers with dense layers with state-of-the-art activation functions (PReLU, Softsign, Softplus) and targeted dropout regularization to improve learning and prevent overfitting. The model is trained using real-world data on Agargaon, Dhaka and gives an RMSE of 8.74 vehicles/interval on roadways and 9.83 at crossroads, with R² = 0.985 (98.5% variance explained). This results in 65–80% error reduction compared to conventional LSTM (RMSE: 15.32) and over 95% compared to classical baseline models such as SVR (19.67), Kalman filter (21.45) and ARIMA (24.91). It is noteworthy that the average inference latency can be less than 50 ms on conventional CPUs, and thus it can be deployed in smart transportation systems with resource constraints.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Md Abdulla Hasan
AU  - Zaffar Abdullah
AU  - Tasnia Noshin Orin
PY  - 2026
DA  - 2026/06/08
TI  - An Optimized Multi-layer LSTM Network for Real-Time Short-Term Traffic Forecasting in Urban Environments
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 1228
EP  - 1243
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_83
DO  - 10.2991/978-94-6239-664-7_83
ID  - Hasan2026
ER  -