Concept Drift Detection in Data Stream Mining: The Review of Contemporary Literature
Mining process such as classification, clustering of progressive or dynamic data is a critical objective of the information retrieval and knowledge discovery; in particular, it is more sensitive in data stream mining models due to the possibility of significant change in the type and dimensionality of the data over a period. The influence of these changes over the mining process termed as concept drift. The concept drift that depict often in streaming data causes unbalanced performance of the mining models adapted. Hence, it is obvious to boost the mining models to predict and analyse the concept drift to achieve the performance at par best. The contemporary literature evinced significant contributions to handle the concept drift, which fall in to supervised, unsupervised learning, and statistical assessment approaches. This manuscript contributes the detailed review of the contemporary concept-drift detection models depicted in recent literature. The contribution of the manuscript includes the nomenclature of the concept drift models and their impact of imbalanced data tuples.
