Dimension Reduction for Big Data Analytics in Internet of Things

Abstract

The number of Internet of Things (IoT) devices continues to grow with the invention of sophisticated applications in smart cities. It is forecasted that there will be 50 billion IoT devices by 2025. These large numbers of IoT devices and sensors are generating a huge amount of data in the various different formats such as plain messages, images, audio, and video. It is important to analyze this large amount of data. However, limited capabilities of IoT devices (such as low-power and computational capability) require efficient and robust methods to deal with the big data analytics. Numerous statistical techniques such as regression analysis, support vector machines, ensembles, decision trees, analysis of variance, correlation and autocorrelation, etc. led to massive amounts of data being processed in novel ways. It is important to reduce the number of variables in data before processing it. Dimension reduction is considered as an effective method to reduce the number of variables in data generated by IoT devices. In this chapter, we first present related work on dimension reduction in IoT systems. Then, we provide a detailed discussion of solutions for dimension reduction with several examples. Finally, we present conclusions and highlight open research areas for data reduction in IoT systems.