4. Feature Selection

In big data analytics, irrelevant and redundant features may not only deteriorate the performances of classifiers, but also slow down the prediction process. Although there is the availability of many classification models for prediction, it is a challenge to choose a set of important features that can lead to a satisfactory classifier. This chapter outlines some achievements of feature selection research in the last decade. Section 4.1 has three subsections. The first is an integrated scheme for feature selection and classifier evaluation in the context of prediction [1]. The second is about two-stage hybrid feature selection algorithms [2]. The third one is the feature selection with attributes clustering by maximal information coefficient [3]. Section 4.2 presents two regularizations for feature selections. They are feature selection with MCP² regularization [4] and feature selection with ℓ _{2, 1 − 2} regularization [5]. Finally, Sect. 4.3 describes two distance-based feature selections. They are the spatial distance join based feature selection [1] and a domain driven two-phase feature selection method based on bhattacharyya distance and kernel distance measurements [6].