A Hybrid Approach Using Unified Rough Set and Clustering Techniques for Generating a Relevant and Significant Feature Subset

Abstract

The advent of information technology and its fast pace has resulted in a continuous stream of data flow posing a major challenge in information extraction and feature selection domains. Machine learning technique, an important tool in handling such large quantities of data has to face the constraints of high dimensionality of data, which has restrained from selecting relevant features to classify patterns. At this juncture, the role of feature selection techniques caught the attention of researchers and a combination of such techniques is attempted on different datasets. In addition to that the potential concept like rough set theory has been integrated, which could bring out the relevance of features in an information system. As a result the data mining concept has become more potent and meaningful. Such approach is attempted in the present work to understand and appreciate the rough set based feature selection techniques involving Fuzzy C-Means and K-Means clustering techniques along with K-Nearest Neighbors to determine outliers. newline