An algorithm for visualization of big data in a two-dimensional space

Abstract

In this paper, a new algorithm for visualization of high-multidimensional data is described. The algorithm follows several steps. At first, centers representing several categories are selected, and Euclidean distances between these centers are calculated in a high-dimensional space. Then these centers are placed in a 2-dimensional space in such a way that distances in this 2-dimensional space are similar to distances in the high-dimensional space. Next individual patterns are placed one-by-one in the 2-dimensional space trying to keep the similar distances in a high-dimensional and 2-dimensional space. With this algorithm, it was possible to visualize many high-dimensional data sets. The algorithm was successfully verified in several real life problems. It turned out that in some cases, which were until now considered as not linearly separable, became easily separable once patterns were transformed in the 2-dimensional space using the proposed algorithm.