Dimensionality reduction using geometric projections: A new technique

Abstract

A new method based on geometric projections for Dimensionality Reduction of multispectral, remotely sensed data is presented. A composite of four different parameters, called the “Clustering Tendency Index” (CTI) has been defined to quantify the suitability of the Dimensionality Reduction methods from the point of view of clustering. The Dimensionality Reduction scheme involves transformation of data from multi-dimensional n-space to a two-dimensional (2D) space, which reduces storage requirements and processing time in addition to facilitating representation in the Cartesian coordinate system. The efficacy of the algorithm is established by experimental studies using different data sets.