AN EFFICIENT INDEXED BASED CLASSIFICATION APPROACH FOR FEATURE EVOLVING DATA STREAMS

Today, rapid growth in hardware technology has provided a means to generate huge volume of data continuously. In most of the real time data stream application data usually reach very rapidly that flows continuously in real time environment .This incoming data streams comprises of several important and interesting patterns underneath. However, mining an essential data out from this data stream, has some major challenges such as infinite length, concept evolution and concept-drift .Earlier studies, carried till now, have been mainly focusing on building accurate classification model. But, keeping glance over the huge amount of classifiers while prediction, require more response time, which makes ensemble approach being impractical for many real-world times critical data stream applications. In order to over this problem, we propose a new enhanced indexing structure that organizes all classifiers of ensemble in order to get fast prediction response in lesser amount of time .In addition to this, ensemble model is updated continuously by integrating new classifiers, while adapting to new trends. Experimental results and theoretical analysis on both real-worlds as well as synthetic data streams exhibit the better performance of our method over the existing techniques.