Designing an inductive data stream management system: the stream mill experience

There has been much recent interest in on-line data mining. Existing mining algorithms designed for stored data are either not applicable or not effective on data streams, where real-time response is often needed and data characteristics change frequently. Therefore, researchers have been focusing on designing new and improved algorithms for on-line mining tasks, such as classification, clustering, frequent itemsets mining, pattern matching, etc. Relatively little attention has been paid to designing DSMSs, which facilitate and integrate the task of mining data streams--i.e., stream systems that provide Inductive functionalities analogous to those provided by Weka and MS OLE DB for stored data. In this paper, we propose the notion of an Inductive DSMS--a system that besides providing a rich library of inter-operable functions to support the whole mining process, also supports the essentials of DSMS, including optimization of continuous queries, load shedding, synoptic constructs, and ...

• Type: Conference
• Year: 2008
• Where: EDBT
• Authors: Hetal Thakkar, Barzan Mozafari, Carlo Zaniolo
• Keywords: Database, EDBT 2008, Light Mining Algorithms, Mining Data, On-line Data Mining