Evaluation of spatio-temporal forecasting methods in various smart city applications

Abstract

Together with the increasing population and urbanization, cities have started to face challenges that hinder their socio-economic and sustainable development. The concept of smart cities, therefore, has emerged during the last years as a response to these problems. Advanced measurement and communication technologies enabled through smart cities have particularly played a key role in dealing with such economic, social and organizational challenges faced during the growing of cities. In this sense, using historical information provided with the mentioned technologies, various forecasting tools have been incorporated into smart city environment in order to manage more effectively its essential components such as smart grids and Intelligent Transportation Systems (ITS). For a further improvement in forecasting accuracy and hence in the management of these smart systems, recently, the information available in space has been also introduced in forecasting tools in addition to that in time. These advanced forecasting approaches, called spatio-temporal methods, have the capability of making use of all the available data collected from different locations. The potential benefits of these approaches have been underlined in various recent studies in the literature. In this paper, a comprehensive overview and assessment of forecasting approaches including both spatial and temporal information have been presented for the purpose of supporting the ongoing efforts for exploiting the available information in smart city applications. With this objective, the spatio-temporal forecasting methods presented in the literature are classified considering their implementation areas and model structures. Furthermore, the similarities and peculiarities of the methods classified are examined in detail, resulted in the compiling of valuable reference information for future studies on improving these approaches.