Concept Drift Challenges in the Internet of Things (IoT) Era of Smart Cities: A Preliminary Investigation

Recently, smart cities and their enabling technologies, such as the Internet of Things (IoT), have grown significantly. This trend has led to an exponential increase in smart devices and, as a result, facing the phenomenon of streaming big data continuously being produced with high volume, speed, and variety. This issue has created an unprecedented opportunity for businesses to exploit this data through a new generation of data-driven applications. However, the real-time, dynamic, non-stationary nature and the changing statistical patterns of big data streams in the IoT present data analysis with concept drift challenges as the leading cause of the gradual decline in efficiency and unreliability of static machine learning models. Dealing with this issue and making real-time decisions in dynamic situations is one of this research field's challenges. Although concept drift and IoT have each been independently researched in the literature, the confluence of these subjects has not yet received any attention. Consequently, this paper marks the initial stride towards tackling the issue in this domain. After discussing the significance and various notions of the concept drift, this paper will highlight some of the latest research works in this field. The paper will present this field's challenges and research opportunities.