Declarative in-network sensor data analysis

Bridging data analysis techniques with classic query processing has long been of interest in the database community. Most approaches, however, are usually developed with a specific domain in mind, e.g. relational, streaming etc., use their own query language, or focus on specific techniques. In this paper, we propose a simple, yet effective, extension to standard or commonly used declarative processing languages to support data mining. Our approach is independent of a particular domain, and by utilizing a query refactoring technique, optimization issues are taken care of by the underlying query processing engine, which is already in place and knows best the setting’s particularities. Therefore, our approach promotes ease of programmability, development, and use of the data mining techniques, with minimal modifications in the query processing stack. We demonstrate our technique through an experimental evaluation, using our prototype system SNEE-A, that runs in-network data analysis given a sensor network deployment, a setting with several critical constraints.