ABSTRACT
This chapter discusses the challenges involved in analyzing streams and provides an overview of the challenges. It also discusses the notions of infinite length and concept drift in streaming data. The chapter describes the aspects of limited labeled data and explores authors’ contributions to the field. Data streams are continuous flows of data being generated from various computing machines such as clients and servers in networks, sensors, call centers, and so on. Most data stream classification techniques concentrate only on the first two issues, namely, infinite length, and concept drift. Traditional data stream classification techniques solve the infinite length problem by providing a one-pass learning paradigm using one of the two approaches: single model or ensemble classification. Concept evolution occurs in data streams when novel classes emerge. Realistic Data Stream Classifier (ReaSC) offers a practical data stream classifier that not only views the data stream classification problem from a real perspective, but also provides a cost-effective solution.
