A framework for distributed data analysis for IoT

Abstract

This chapter introduces a framework for distributed data analysis based on the core idea Fog computing to use local resources to reduce the overhead of centralized data collection and processing. This is achieved by learning local models of the data at the nodes, which are then aggregated to construct a global model at a central node. This chapter explains how clustering algorithms enable the central node to handle nonhomogeneity in the data collected at different nodes. It then describes an efficient incremental modeling technique, which facilitates the calculation of local models in highly resource constrained nodes. This chapter also provides experimental results to demonstrate the benefi..