Editorial message: special track on data streams

Abstract

Advances in data acquisition hardware and embedded systems have led to the data stream era. A growing number of emerging applications varying from business to scientific to industrial ones continuously generate open-ended data streams. In practice, such data cannot be stored but must be both queried and analyzed as they arrive, discarding it right away. In many cases, we need to extract some sort of knowledge from these continuous streams that challenge the scalability of several batch-learning methods. Therefore, this new field has attracted researchers from different disciplines over the past few years. Examples of data streams include customer click streams, networks event logs, telephone records, large sets of web pages, multimedia data, scientific data, and sets of retail chain transactions. Applications include credit card fraud protection, target marketing, and intrusion detection, for which it is not possible to collect all relevant input data. In these environments, KDD systems have to operate online under memory and time limitations.