Adaptive ensemble based stream data learning model for handling imbalanced and concept drift

Abstract

With the advancement of information technology, organizations newlinetend to generate a tremendous amount of high-velocity data streams. The newlinestaggering growth of such extensive volume time-changing data gives rise to newlinenumerous critical issues and constraints such as concept drift and class newlineimbalance on the design of learning algorithms. The concept drift is one of the newlinechallenging issues, and recognizing the sequential patterns over continuously newlineevolving data streams are even more daunting. Nowadays, the ensemble newlinelearning model has gained in significance as they incrementally learn the newlinecontinuous flow of data for ensuring rapid response to the concept drifts. newlineBesides, it easily adapts quickly to both gradual and sudden concept drifts in newlinethe real-time data streams. On the other hand, the sampling techniques are newlinebroadly used to handle the data streams with the imbalanced class distribution. newlineHowever, the concept drift and class imbalance in data streams significantly newlinehinder the performance of the learning algorithms, and the issue becomes newlineextremely challenging when they happen simultaneously. Also, the concept newlinedrift detection methods are sensitive to the imbalanced class and turn into less newlineefficient while dealing with a higher degree of imbalanced data. To cope up newlinewith these issues, this research work offers the two significant contributions in newlinestream data environment such as Handling Imbalanced Data with Concept newlineDrift (HIDC), and Stream data mining On the fly using Adaptive online newlinelearning Rule model (SOAR). newlineThe initial contribution focuses on incoming massive imbalanced newlinedata with concept drift. newline