A Hybrid Approach for Data Analytics for Internet of Things

TL;DR

This paper proposes a hybrid data analytics approach for IoT that combines local feature extraction on edge devices with cloud processing to enhance privacy, reduce network load, and maintain accuracy.

Contribution

It introduces a hybrid model integrating edge and cloud analytics, utilizing data fusion for feature extraction on resource-constrained devices.

Findings

Reduced data transmission by local feature extraction

Maintained high accuracy in data analysis

Enhanced privacy through local data processing

Abstract

The vision of the Internet of Things is to allow currently unconnected physical objects to be connected to the internet. There will be an extremely large number of internet connected devices that will be much more than the number of human being in the world all producing data. These data will be collected and delivered to the cloud for processing, especially with a view of finding meaningful information to then take action. However, ideally the data needs to be analysed locally to increase privacy, give quick responses to people and to reduce use of network and storage resources. To tackle these problems, distributed data analytics can be proposed to collect and analyse the data either in the edge or fog devices. In this paper, we explore a hybrid approach which means that both innetwork level and cloud level processing should work together to build effective IoT data analytics in order to overcome their respective weaknesses and use their specific strengths. Specifically, we collected raw data locally and extracted features by applying data fusion techniques on the data on resource constrained devices to reduce the data and then send the extracted features to the cloud for processing. We evaluated the accuracy and data consumption over network and thus show that it is feasible to increase privacy and maintain accuracy while reducing data communication demands.