Araya, Kibrom Desta (Nara Institute of Science and Technology), Arai, Ismail (Nara Institute of Science Technology)

Evaluating Transformer Models for Road Traffic Volume Forecasting with Weather-Aware Inputs

Scheduled for presentation during the Invited Session "S24a-Traffic Control and Connected Autonomous Vehicles: benefits for efficiency, safety and beyond" (TH-LM-T24), Thursday, November 20, 2025, 11:10−11:30, Coolangata 3

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

Transformer models, originally developed for natural language processing, have recently gained attention in time series forecasting. In this study, we extend their application to road traffic volume prediction as part of efforts to mitigate urban congestion. We utilize five years of sensor data collected from three major traffic junctions in Istanbul, Turkey, and evaluate the performance of several Transformer-based models, including Informer, Autoformer, Reformer, and the recently proposed iTransformer. Our experiments show that the iTransformer consistently outperforms other Transformer variants and also surpasses our baseline models of LSTM and a simple linear models. Additionally, we investigate the effect of incorporating weather information, specifically temperature and categorical weather conditions, on forecasting performance. Across the three junctions, our best-performing models achieved mean absolute errors ranging from 0.107 to 0.284 for short-term forecasts over an 8-step prediction timesteps.