Handling of Recurrence Concept Drift in Data Stream using Timestamp of Auxiliary Learning Model

Data stream is a collection or sequence of data instances of infinite length. Stream classification or online classification is more challenging task due to speed, diversity of concept or nature, type of distribution (linear or skewed), heterogeneous of data sources, lack of rereading of instances and possibility of recurrence. This paper focuses on the concept drift under recurrence. The major challenge in data stream is handling of high volume of data of infinite length. Classification of instances under concept drift and recurrence is more difficult due to maintenance of past classifier results. To handle this situation in more efficient manner through swapping technique followed operating system’s demand