An approach for adaptive learning of imbalanced and concept drifted data streams

Abstract

newline Over the past few decades, all sectors like automobile, newlineconstruction, entertainment, banking, agriculture, medicine, etc. have newlinebeen experiencing digital transformations. In many real-world newlineapplications, the continuous flow of tremendous digital data is rapidly newlinegenerated. The learning of such data streams is highly essential to newlineextract the knowledge from them. The necessity of one-time processing newlineof the high-speed, boundless stream of data makes data stream mining a newlinechallenging task. newlineData streams integrate dynamicity because of the newlinenonstationary environment in which data samples may experience the newlineclass imbalance and concept drifts. The unequal proportion of class-wise newlineinstances in data is called a class imbalance. However, the variations in newlinedata distribution over the period are defined as concept drifts. newlineResearchers seldom discuss a combined method to handle the class newlineimbalance and concept drifts in dynamic data streams. The current newlineresearch work addresses the challenge of adaptive learning of nonstationary binary data streams exhibiting both class imbalance and newlineconcept drift together. newlineThe presented research work contributes to a passive drift newline