Data stream management systems for processing and mining RFID streams

摘要

RFID technology offers significant advantages over traditional object-tracking technology and is increasingly adopted and deployed in many applications. Such applications generate large volumes of streaming data, which have to be filtered, processed, and transformed into higher-level information for integration into business applications. For instance, RFID data are highly temporal and form diverse patterns of complex temporal events that must be detected and managed in different applications. This situation calls for the development of powerful Data Stream Management Systems (DSMS) that can support complex event processing of massive RFID data streams with Quality of Service (QoS) guarantees. Such DSMS must enable applications ranging from low-level data cleaning, to complex temporal event pattern detection and sophisticated data mining on RFID data streams. This talk will (i) describe how complex event detection and management can be achieved through Kleene-closure extensions of continuous query languages, and (ii) show that data stream mining can be achieved through user-defined aggregates written in the same language. This advanced functionality is realized efficiently in Stream Mill-a DSMS designed for power and extensibility that provides a powerful platform for processing RFID information.