A model-driven framework for data quality management in the Internet of Things

Abstract

The internet of Things (IoT) is a data stream environment where a large scale deployment of smart things continuously report readings. These data streams are then consumed by pervasive applications, i.e. data consumers, to offer ubiquitous services. The data quality (DQ) is a key criteria for IoT data consumers especially when considering the inherent uncertainty of sensor-enabled data. However, DQ is a highly subjective concept and there is no standard agreement on how to determine “good” data. Moreover, the combinations of considered measured attributes and associated DQ information are as diverse as the needs of data consumers. This introduces expensive overheads for developers tasked with building DQ-aware IoT software systems which are capable of managing their own DQ information. To effectively handle these various perceptions of DQ, we propose a Model-Driven Architecture-based approach that allows each developer to easily and efficiently express, through models and other provided resources, the data consumer’s vision of DQ and its requirements using an easy-to-use graphical model editor. The defined DQ specifications are then automatically transformed to generate an entire infrastructure for DQ management that fits perfectly the data consumer’s requirements. We demonstrate the flexibility and the efficiency of our approach by generating two DQ management infrastructures built on top of different platforms and testing them through a real life data stream environment scenario.