An ontology based approach to intelligent data mining for environmental virtual warehouses of sensor data

The rapid development of sensor technology led to large deployments of collections of sensing nodes working together to collect information for light, temperature, atmospheric pressure, air pollutions parameters, images, and other relevant data according to specific Environmental applications. The ability of the sensor networks to collect information accurately and reliably enables building both real-time detection and early warning systems. Environment parameter measurement and supervision received a lot of attention from both academic and industrial research communities around the world. Applications supporting decisions for rapid coordinated responses to threats such as bushfires, tsunamis, earthquakes, and other crisis situation have been devised. In this paper a combined ontology and rule-bases intelligent system build over a platform for sensor data warehousing for decision support is introduced. A combined ontology of sensor networks and sensor data for environment parameter measurement is designed and discussed. A set of rule-based systems each tailored to a domain of networked sensors are linked together through a middle-layer of rule-base systems which reason with facts classified and structured as instances of the common sensor ontology. A set of meta-rules direct the intelligent activity of the whole system and requires individual rule-based systems to reason upon specific facts. The paper will also discuss the architecture and algorithms for data mining, fusion and reasoning for specific situation in Environment monitoring and supervision. Results obtained from experiments performed on the system prototype are given at the end of this paper.