You are in the accessibility menu

Please use this identifier to cite or link to this item: http://acervodigital.unesp.br/handle/10400.2/3873
Title: 
Stock market series analysis using self-organizing maps
Author(s): 
Abstract: 
  • In this work a new clustering technique is implemented and tested. The proposed approach is based on the application of a SOM (self-organizing map) neural network and provides means to cluster U-MAT aggregated data. It relies on a flooding algorithm operating on the U-MAT and resorts to the Calinski and Harabask index to assess the depth of flooding, providing an adequate number of clusters. The method is tuned for the analysis of stock market series. Results obtained are promising although limited in scope.
  • Neste trabalho é implementada e testada uma nova técnica de agrupamento. A abordagem proposta baseia-se na aplicação de uma rede neuronal SOM (mapa auto-organizado) e permite agrupar dados sobre a matriz de distancias (U-MAT). É utilizado um algoritmo de alagamento ("flooding") sobre a U-MAT e o índice de Calinski e Harabasz avalia a profundidade do alagamento determinando-se, assim, o número de grupos mais adequado. O método é desenhado especificamente para a análise de séries temporais da bolsa de valores. Os resultados obtidos são promissores, embora se registem ainda limitações.
Issue Date: 
2014
Citation: 
Matos, Diogo; Marques, Nuno C.; Cardoso, Margarida G. M. S. - Stock market series analysis using self-organizing maps. "Revista de Ciências da Computação" [Em linha]. ISSN 1646-6330 (Print) 2182-1801 (Online). Vol. 9, nº 9 (2014), p. 79-90
Publisher: 
Universidade Aberta
Keywords: 
  • Financial markets
  • SOM
  • Clustering
  • U-Matrix
  • Flooding
  • Neural networks
  • Mercados financeiros
  • Agrupamento
  • Alagamento
  • Redes neuronais
  • REA
URI: 
Rights: 
openAccess
Type: 
outro
Source:
http://repositorioaberto.uab.pt/handle/10400.2/3873
Appears in Collections:Recursos Educacionais Abertos (REA) - Universidade Aberta de Portugal

There are no files associated with this item.
 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.