Evaluation of Dimensionality Reduction Techniques for Big data

Abstract

In this digital era, big data has very high dimension and requires large amount of space for its data storage. Hence a lossless data interpretation will be difficult when big data contains large dimension. But, all these dimensions in big data may not be relevant or they may be interrelated and hence redundancy may exist in attribute set. Dimensionality reduction is a technique which focusses on downsizing the attributes and complication of a high dimensional data. In this paper, a detailed study of different dimensionality reduction techniques namely principal component analysis (PCA), linear discriminant analysis (LDA), kernel principal component analysis (KPCA), singular value decomposition (SVD), independent component analysis (ICA) has been proposed. Furthermore, it also provides comparative analysis based on various parameters.