Decentralized Spatial Data Mining for Geosensor Networks
Patrick Olivier Laube, Matt Duckham · Chapman & Hall/CRC data mining and knowledge discovery series · 2009
Advances in distributed sensing and computing technology offer new, reliable, and costeffective means to collect fine-grained spatiotemporal data. Conventional spatiotemporal data mining procedures, however, are based on centralized models of information processing, where sophisticated and powerful central systems collate and process global information. By contrast, decentralized spatial computing systems require new techniques for in-network knowledge discovery. This chapter introduces the notion of decentralized spatial data mining, where individual sensor-enabled computing nodes possess only local knowledge about their immediate neighborhood, but derive global knowledge through local collaboration and information exchange. The chapter then presents four strategies for decentralized spatial data mining, illustrating the concept of decentralization with three simple decentralized algorithms for the classical spatial data mining task of clustering.