Multi-modal data fusion and explainable artificial intelligence for smart traffic flow prediction in urban mobility networks

Urban mobility networks encompass a wide range of populations, vehicles, cities, etc. that experience mobility for specific purposes. Such elements in an urban network leave out traces of spatio-temporal data from which tracking, localization, and information sharing are eased. However, predicting traffic flows is crucial due to irregular data and complex computations. To provide a solution to such a problem, this article introduces a multi-functional method by assimilating multi-modal data fusion (MMDF) with explainable artificial intelligence (EAI). The joint method improves traffic prediction by reducing error variants based on mobility management and traffic representations. This is accomplished using three different functions: classified data fusion, traffic prediction using explainable AI, and prediction-based mobility management. These functions are consolidated by utilizing spatial and temporal data that considers the impact of different factors such as population, vehicles, etc. This methodology improves the traffic flow prediction by 14.53% by reducing the error by 14.91% for the varying intervals.