Data Reduction Techniques in Fog Data Analytics for IoT Applications

Internet of Things (IoT) has been changing the way of day-to-day life activities because of the use of sensors, wearables, and smartphones. The connected devices form the basis for smart systems such as smartwatches, smart cities, smart cars, smart grids, and smart buildings. The devices are capable of collecting large quantities of data. The data once collected need to be analyzed intelligently to leverage the potential of the data. The existing analytical approach of carrying out analytics lies in the principles of data transfer to cloud for storage and processing, subsequently delivering the results to the IoT-based software applications. Fog Computing (FC) has paved the way for providing an alternate way for IoT data analytics compared to the centralized cloud computing approach for analytics. FC is a paradigm based on the approach of carrying out computation and analytics at the edge devices rather than the cloud. However, the latency analysis in FC remains a challenge. In this chapter, the data reduction techniques are explored with FC for IoT applications. The techniques included are Missing Values Ratio, Low Variance Filter, High Correlation Filter, Principal Component Analysis (PCA), Random Forrest/Ensemble Trees, Backward Feature Elimination, and Forward Feature Construction. A case study with PCA method for Fog Data Analytics is discussed at the end of the chapter. It demonstrates the effectiveness of data reduction methods in FC.