Using nonlinear dimensionality reduction techniques in big data analysis

Abstract

In recent years, the huge development in the measure of data has been noted. This becomes a first step of the big data. Big data can be defined as high volume, velocity and variety of data that require a new high-performance processing. The reduction of dimensions is one of the most important methods that are necessary in the field of big data analysis. Ideally, the reduced representation should have a dimensionality that corresponds to the intrinsic dimensionality of the data. There are two important procedures, the first one is dimensions reduction, and the second one is putting the data into its model and then estimating it. In this paper, we use two techniques of nonlinear dimensional reduction. The first one includes kernel principal components analysis (KernelPCA) along with modified kernel principal components analysis smooth (KernelPCAS) and the second one is neural network. The mean square error (MSE) was used to demonstrate the effectiveness of nonlinear reduction methods in the analysis of big data as a resourceful tool for this purpose.