Multiscale spatio temporal analysis for traffic prediction in smart cities

Abstract

Recent advancements in computing technologies and the availability of location specific data has inspired the development of Intelligent Transport Systems for quick adaptation in smart cities. Urban sprawl and rise of economy has triggered travel aspirations among the inhabitants. Meanwhile, city infrastructure management and traffic control authorities are posed with the challenges of ensuring hassle free commutation within the city. Recently, traffic prediction has received keen interest and is vital to provide accurate and timely information for travel planning and minimise operational costs. Modernization of cities with sophisticated amenities for generating urban data has facilitated the collection and storage of multivariate time series data with geo-tagged information from transportation sector and meteorological stations. Initially, traffic prediction methods were based on statistical techniques and eventually leap forwarded to data driven approaches to handle stochastic and dynamic traffic. Nevertheless, traffic prediction is highly challenging with the presence of complex factors and hidden information. Moreover, a majority of literatures have focused on time series traffic prediction, ignoring the crucial components of spatio-temporal domain and external factors that impact traffic prediction. newline