A Clustering-Based Approach to Reduce Feature Redundancy

Research effort has recently focused on designing feature weighting clustering algorithms. These algorithms automatically calculate the weight of each feature, representing their degree of relevance, in a data set. However, since most of these evaluate one feature at a time they may have difficulties to cluster data sets containing features with similar information. If a group of features contain the same relevant information, these clustering algorithms set high weights to each feature in this group, instead of removing some because of their redundant nature. This paper introduces an unsupervised feature selection method that can be used in the data pre-processing step to reduce the number of redundant features in a data set. This method clusters similar features together and then selects a subset of representative features for each cluster. This selection is based on the maximum information compression index between each feature and its respective cluster centroid. We present an empirical validation for our method by comparing it with a popular unsupervised feature selection on three EEG data sets. We find that our method selects features that produce better cluster recovery, without the need for an extra user-defined parameter.