Resumen de Enhancing self-organizing map capabilities with graph clustering: : An application to financial markets
This work introduces an algorithm combining self-organizing maps (SOMs) with graph filtering procedures and discusses how to employ it for analysing financial markets. This method defines graphs resulting from SOM leader units and the calculation of the correlation among them. The technique is applied on financial datasets, with an eye to markets at different levels of (in)stability; they were analysed both on a period of 1 year and with a moving window along a time span of 10 years, comparing graph topologies. In both cases, enhancing SOMs with graphs made it possible to emphasize relations among enterprises and to check for the emergence of critical patterns
