Traffic and Weather Data Fusion for Traffic Prediction in Sustainable Cities

Anticipating traffic patterns is a vital component in efficiently managing traffic within smart cities. This proactive approach contributes to alleviating traffic congestion, promoting environmental preservation, reducing commute times, and strengthening overall safety measures. Numerous studies have verified that external factors, such as weather conditions, exert an influence on traffic patterns. Therefore, utilizing these factors as supplementary variables can improve traffic prediction accuracy. This paper presents a novel prediction model called the multi-input sequential multihead attention (MI-SMHA) model. This model integrates weather condition information into traffic prediction tasks, aiming to enhance prediction accuracy and computational efficiency. It utilizes advanced techniques from sequential modeling and attention mechanisms, specifically tailored to handle traffic and weather data such as temperature, wind speed, precipitation, visibility, and humidity. This integration aims to leverage the complementary nature of weather conditions in forecasting traffic patterns, yet it remains flexible enough to be generalized to support a wide range of multivariate time series prediction tasks. Data from real-life traffic detectors are utilized to perform experiments and comparisons with two baseline models and three state-of-the-art models to validate and assess the efficiency of the proposed model. The MI-SMHA model was efficient and reliable in forecasting future traffic flow, significantly reducing errors compared to the other models.