Special Issue "Data Stream Mining and Soft Computing Applications"
Sprache des Titels:
Englisch
Original Kurzfassung:
In current industrial systems, the necessity of data stream mining and learning from data streams is increasingly becoming more prevalent and urgent, due to speed, volume and on-line nature of the data generated by such systems. Especially in Big Data, huge data bases and cloud computing environments, there is an intrinsic demand to perform mining and modeling from data in an efficient and economic manner. Therefore, batch off-line approaches are often too time and memory intensive, and cannot process the data at the high enough rate that is often desired (e.g., as in case of fast on-line stream processing environments). This is true even when batch and off-line approaches are applied to sliding windows or onto streaming samples gathered from reservoir computing techniques, because they usually do not operate in a single-pass, incremental manner, but require iterative optimization and learning procedure over the complete data set --- an unrealistic scenario in Big Data and cloud computing environments as data is always that huge that it cannot be loaded into the virtual memory at once.