Self Organising Maps: An Interesting Tool for Exploratory Data Analysis

The Self-Organizing Map (SOM) is an unsupervised neural network based on competitive learning. It projects highdimensional input data onto a low dimensional (usually two-dimensional) space. Because it preserves the neighborhood relations of the input data, the SOM is a topology-preserving technique. A topographic map is a twodimensional, nonlinear approximation of a potentially high-dimensional data manifold, which makes it an appealing instrument for visualizing and exploring high-dimensional data. The Self-Organizing Map (SOM) is the most widely used algorithm, and it has led to thousands of applications in very diverse areas. In this study, we will introduce the SOM algorithm, discuss its properties and applications, and also discuss some of its extensions and new types of topographic map formation, such as the ones that can be used for processing categorical data, time series and tree structured data.