Deep Embedding Data Fusion Scheme Using Variational Graph Auto-Encoder in IoT Environments

Abstract

Internet of things (IoT) is a promised paradigm for developing smart systems and architectures. IoT is a framework where everyday objects can be equipped with capabilities for identifying, sensing, networking and processing that will allow them to interact over the Internet with each other and with other devices and services to achieve some objectives. In IoT, the sensing devices can collect and manage data that exist in their surrounding environments. So, data fusion problem is a major challenge for the future in order to allow highly effective, reliable and accurate management and decision-making of IoT environments. To meet this challenge, this paper introduces a new data fusion scheme uses a variational graph auto-encoder as a deep learning method for creating deep embedding data graph. This graph represents the similarities among data items which can be divided into groups that can be fused together to minimize the amount of transferred data and latency time of the collected data to the cloud server and the energy consumption by IoT devices. The conducted simulations experiments and results show that the proposed scheme can achieve a reasonable performance in terms of latency time, energy consumption, and data fusion accuracy compared to non-fusion schemes.

Tasks

Reproductions