Improving Quality of Data: IoT Data Aggregation Using Device to Device Communications

Abstract

With the widespread adaptation of the Internet of things (IoT), we are already witnessing a deluge of IoT data analytics applications. IoT data analytics can be defined as a process to control and optimize decision making in real time, by analyzing huge chunks of IoT sensor data. Moreover, data analytics stands on the shoulders of sensor data aggregation that includes data pre-processing and routing. This paper envisions two open problems of data aggregation, first, raw IoT sensor data is highly uncertain, second, the traditional algorithms are not fit for processing highly uncertain sensor data. This is formally known as data veracity problem. This paper proposes a data aggregation scheme for highly uncertain raw IoT sensor data collected using the device to device communication. The approach initially reconstructs the subspace using sample data and then it iteratively tracks down the low-rank approximation of the dominant subspace in the presence of high uncertainties at the fog server. Later, the robust dominant subspace is used to estimate a more reliable true sensor data matrix from the highly uncertain raw IoT sensor data traffic matrix. Moreover, the proposed scheme achieves the aforementioned tasks while processing the raw IoT sensor data without any prior information, i.e., in a fully unsupervised fashion. The existing literature based on sampling, approximation, and data reduction either causes random data reduction or destruction of global characteristics of the raw data. However, unlike the existing solutions, the proposed method removes the uncertainties while preserving the global characteristics of the raw data. Performance evaluations conducted using both the real world sensor data and synthetic data injected with noise, outliers, and missing values. Experimental results show that the proposed approach can estimate a reliable true sensor data matrix in the presence of high uncertainties and the energy efficient device to device communication-based data delivery mechanism can accommodate a large number of IoT devices.