Significant applications require data stream mining algorithms to run in resource-constrained environments. Thus, adaptation is a key process to ensure the consistency and continuity of the running algorithms. This chapter provides a theoretical framework for applying the granularity-based approach in mining data streams. Our Algorithm Output Granularity (AOG) is explained in details providing practitioners the ability to use it for enabling resource-awareness and adaptability for their algorithms. Theoretically, AOG has been formalized using the Probably Approximately Correct (PAC) learning model allowing researchers to formalize the adaptability of their techniques. Finally, the integration of AOG with other adaptation strategies is provided.
|Title of host publication||Foundations of Computational Intelligence (6)|
|Place of Publication||Berlin / Heidelberg|
|Number of pages||20|
|Publication status||Published - 2009|
|Name||Studies in Computational Intelligence|