Adaptive stream mining: Pattern learning and mining from evolving data streams.

*(English)*Zbl 1204.68153
Frontiers in Artificial Intelligence and Applications 207. Amsterdam: IOS Press (ISBN 978-1-60750-090-2/hbk; 978-1-60750-472-6/ebook). xii, 212 p. (2010).

The book targets two current challenges in data mining in what concerns incoming data streams, i.e., time-changing massive data and closed tree patterns, and eventually treats the combination of both. In the real-world, fast environment of the digital era, data arrive at a very high speed and must be mined within strict constraints of space, time and sample size. Under these circumstances, the author first introduces a framework for algorithms that adaptively learn from evolving and sizeable data streams that is endowed with an adaptive sliding window algorithm. The work then continues with a formal study of trees from the perspective of closure-based mining. It eventually combines the results to design methods for mining closed trees from dynamic data streams. The book is the published PhD thesis of the author and as a consequence it puts forward interesting and modern ideas in the current trend of research in the field, with achievements both on the theoretical and practical levels and with the accomplished aim of shifting towards green computing.

Reviewer: Catalin Stoean (Craiova)