A Detailed Study on Ensemble Learning Techniques on Predicting Novel Class For Outlier Data

In traditional data mining based learning approaches, learning of exploring and evolving data streams produces the various challenges in terms of continuous stream of data, unbounded data and high Speed data characteristics from the various data repositories. In order to handle those difficulties, many scalable and effective learning models have been employed for data classification after feature extraction and reduction process. However default classifier requires effective learning model to classify the constantly evolves data over time on data streams. In this paper, a detailed study has been carried out on ensemble learning techniques for classifying the data streams. Those analysing model cope will consider the variation of the data in term of concept drift and feature drift. Further those drift data has been analysed on basis of concept and semantic of data. Moreover features of data which evolve will be efficiently handled on ensemble classification model. Finally performance of the models has been evaluated on data evolved as error driven representativeness learning along various constrained on the classification through computation of association of the feature and its weight on adaptive sliding windows of the data streams.