A review on concept evolution technique on data stream

In Recent years data stream classification has been an extensively studied research problem. Data streams are continuous and rapid flow of data. Data streams include Call center records, network traffic data , sensor data and so many other data. Main problem of data streams is its infinite length, concept drift temporal behavior, concept evolution and feature evolutions. It is impractical to store the historical data for training, Because the data streams which consists the historical data are infinite in length. There is a lot of work done on the existing challenges such as concept drift and infinite length, But less concentrated towards Concept evolution. when the new classes or novel classes are invoking in data streams, this scenario is called concept evolution. As we know the existing challenges are concept drift and infinite length, we address concept evolution detection in this paper. In this paper, enhance approach is used for detection of unseen classes in data stream using adaptive outlier detection, discrete Gini coefficient and multiple unseen classes detection.