Multivariate data classification using PolSOM

Polar self-organizing map (PolSOM), a novel data visualization algorithm, projects data on a polar map with two variables, radius and angle, which represent data weight and feature respectively. Compared with self-organizing map (SOM), which is a traditional method for dimensionality reduction and data classification, PolSOM visualizes not only the inter-neuron distance, but also the differences among clusters in terms of weight and feature. PolSOM sets each neuron as a benchmark to group the similar data together, and reflects the data characteristic by their polar coordinates. In this paper, two multivariate data sets are provided to demonstrate the performance of PolSOM. All simulations are compared with SOM and ViSOM.