Using Self Organising Maps to Visualize Large Multivariate Datasets

📄 Project Abstract

Visual Analytics is an emerging area of study to leverage human cognitive abilities in reasoning and understanding patterns that may exist in complex data. In particular the study of Visual Analytics attempts to maximise the extraction of relationships among data items through software that enables analytical reasoning through the visualization of large multivariate data sets. The Self-organising map (SOM) is one such technology which allows data to be clustered into areas of similarity, and provides a variety of visualizations to enable further reasoning about relationships in the underlying data from the properties of these emergent clusters. The SOM was introduced by Kohonen in 1981, as an extension to unpublished work dating back to 1976 [33], and uses a neural network to automatically cluster similar data items using vector quantization and competitive learning to create an efficient classifier for new input data, as vectors of constituent attributes. An application that creates S...