Abstract
Dimensionality reduction of a data set by selecting or extracting relevant and nonredundant features is an essential preprocessing step used for pattern recognition, data mining, machine learning, and multimedia indexing. Among the large amount of features present in real life data sets, only a small fraction of them is effective to represent the data set accurately. Prior to analysis of the data set, preprocessing the data to obtain a smaller set of representative features and retaining the optimal salient characteristics of the data not only decrease the processing time but also lead to more compactness of the models learned and better generalization. In this regard, a novel dimensionality reduction method is presented here that simultaneously selects and extracts features using the concept of feature significance. The method is based on maximizing both relevance and significance of the reduced feature set, whereby redundancy therein is removed. The method is generic in nature in the sense that both supervised and unsupervised feature evaluation indices can be used for simultaneously feature selection and extraction. The effectiveness of the proposed method, along with a comparison with existing feature selection and extraction methods, is demonstrated on a set of real life data sets.