A Framework for Mining High Dimensional Data for Feature Subset Selection?

Features are representative characteristics of data sets. Identifying such features in a high dimensional dataset play an important role in real world applications. Data mining is best used to determine important features. Selecting important features from a subject of identified features can help in making expert decisions. However, efficient identification of such feature subset and selection is a challenging problem. Recently Song et al. proposed a solution that is capable of selecting subset of features with good quality. They used clustering approach before selecting representative features for final selection. Similar work is carried out in this paper which demonstrates the proof of concept. The proposed solution makes use of clustering for achieving the goal of the system. The empirical results reveal that the application is useful. The results are compared with many existing algorithms like C4.5, Naïve Bayes, IB1 and RIPPER.