Building a New Classifier in an Ensemble Using Streaming Unlabeled Data

The GC3 framework : grid density based clustering for classification of streaming data with concept drift.

“…In [32], Woo et al developed an ensemble approach to classifying streaming data based on misclassified sample points. In this paper and in the following work in [33] and [34], they have demonstrated the effectiveness of their algorithm which is based on clustering and subsequent classification of streaming data.…”

“…In [32] the entire testing data was considered to be labeled, however in [33] and [34] it is shown that similar performance can be obtained by using only partially labeled data.…”

“…It also provides a brief overview of the existing techniques developed to deal with concept drift. The ensemble classification techniques proposed by Woo et al in [32,33,34] and the grid density clustering described in [35,36] lay the foundation for the development of the GC3 framework. Chapter 3 introduces the GC3 framework and provides detailed explanation of its various components and their working.…”

“…One of the most widely used techniques for dealing with streaming data is the use of an Ensemble of Classifiers for modeling the stream [21][22][23][24][25][26][27][28][29][30][31][32][33][34]. The basic principle of ensemble classifiers is that of combining several weak and independent classifiers to produce a strong model on the entire data.…”

“…This method is due to Woo et al [32][33][34]. In this approach, an ensemble maintains a set of models, each native to a particular region in the multidimensional feature space.…”