FoT-Stream - Source codes and dataset used to evaluate a Fog Platform for Data Stream Analytics in IoT

Our experiments were designed to deal with the problem of processing and analyzing data streams, produced by IoT devices, in Fog Computing. In our research, we adapted in Fog environments two well-known techniques widely used in the Signal Processing area: Wavelet transform and Concept Drift detection. Wavelet transform was responsible to decompose the data streams into a set of components capable of keeping the general data behavior without using all observation values. Complementary, Concept Drift detection was adopted to reduce the data transmission on the network, which only happens when the general behavior changes over time. Therefore, our research provides three contributions: i) a smaller amount of data transmitted by the network; ii) online modeling by detecting any temporal changes; and iii) full-time connectivity to the Internet is no longer required in IoT context.