Mathematical Methods for Data Fusion in IoT: A Survey

IoT (Internet of Things) is a paradigm that connects multiple and diverse smart objects via internet, it is currently widely used in a number of real-life fields. Smart interconnected devices generate tremendous amounts of raw data using their sensors. Such data need to be processed, analyzed, and mined efficiently to obtain improved information that help to take corrective actions and provide advanced services. This paper provides a critical review of existing methods on data fusion in IoT context with a particular focus on the most applicable and mature mathematical methods. Inherent characteristics of data fusion theories are restated, then a comparative study is conducted by pointing out the advantages and limitations of each method taking into account data imperfections (uncertainty, inaccuracy, etc.) and IoT environments constraints (real-time, etc.). This work constitutes a big step towards building a guide for researchers working in the field of data fusion in IoT.