A Data Traffic Reduction Approach Towards Centralized Mining in the IoT Context

Abstract

The use of Internet of Things (IoT) technology is growing each day. Its capacity to gather information about the behaviors of things, humans, and process is grabbing researchers’ attention to the opportunity to use data mining technologies to automatically detect these behaviors. Traditionally, data mining technologies were designed to perform on single and centralized environments requiring a data transfer from IoT devices, which increases data traffic. This problem becomes even more critical in an IoT context, in which the sensors or devices generate a huge amount of data and, at the same time, have processing and storage limitations. To deal with this problem, some researchers emphasize the IoT data mining must be distributed. Nevertheless, this approach seems inappropriate once IoT devices have limited capacity in terms of processing and storage. In this paper, we aim to tackle the data traffic load problem by summarization. We propose a novel approach based on a grid-based data summarization that runs in the devices and sends the summarized data to a central node. The proposed solution was experimented using a real dataset and obtained an expressive reduction in the order of 99% without compromising the original dataset distribution’s shape.