Forecasting economic time series using chaotic neural networks

• Victor H. Gonçalves USP
• João L. G. Rosa USP

DOI: https://doi.org/10.5753/eniac.2018.4470

Resumo

This paper describes the application of KIII, a biologically more plausible neural network model, for forecasting economic time series. K-sets are connectionist models based on neural populations and have been used in many machine learning applications. In this paper, this method was applied to IPCA, a Brazilian consumer price index surveyed by IBGE. The values ranged from August 1994 to June 2017. Experiments were performed using four non-parametric models and seven parametric methods. The statistical metric RMSE was used to compare methods performance. Freeman KIII sets worked well as a filter, but it was not a good prediction method. This paper contributes with the use of non-parametrics models for forecasting inflation in a developing country.

Referências

(2018a). 38160 reais =. Dollar values obtained by conversion of Brazilian real to US dollar made by Google calculator on 07/03/2018.

(2018b). 954 reais =. Dollar values obtained by conversion of Brazilian real to US dollar made by Google calculator on 07/03/2017.

Brazilian Central Bank (2017a). Broad national consumer price index (ipca). Time Series Management System - v2.1.

Brazilian Central Bank (2017b). Time series management system - v2.1.

Chang, C.-C. and Lin, C.-J. (2016). Libsvm – a library for support vector machines.

Ferrero, C. A. (2009). Algoritmo knn para previs˜ao de dados temporais: funções de previs˜ao e critérios de seleção de vizinhos próximos aplicados a variáveis ambientais em limnologia. Master’s thesis, Instituto de Ciências Matemáticas e de Computação, Universidade de S˜ao Paulo.

Freeman, W. J. (1975). Mass Action in the Nervous System. Academic Press New York, San Francisco, London.

Gooijer, J. G. D. and Hyndman, R. J. (2006). 25 years of time series forecasting. International Journal of Forecasting, 22(3):443–473.

Harvey, A. (1985). Trends and cycles in macroeconomic time series. Journal of Business & Economic Statistics, 3(3):216–227.

Instituto Brasileiro de Geografia e Estatística (2017). Sistema nacional de Índices de preços ao consumidor ipca e inpc.

Kozma, R. and Beliaev, I. (2004). Time series prediction using chaotic neural networks: case study of ijcnn cats benchmark test. In Proc. 2004 IEEE international joint conference on neural networks, volume 2, pages 1609–1613.

Kozma, R. and Freeman, W. J. (2001). Chaotic resonance - methods and applications for robust classification of noisy and variable patterns. International Journal of Bifurcation and Chaos, 11(6):1607–1629.

Kozma, R., Piazentin, D. R. M., and Rosa, J. L. G. (2013). Cognitive clustering algorithm for efficient cybersecurity applications. In 2013 International Joint Conference on Neural Networks (IJCNN), pages 1–8.

Li, H. and Kozma, R. (2003). A dynamic neural network method for time series prediction using the kiii model. In Proc. 2003 IEEE International Joint Conference on Neural Networks, pages 347–352.

Parmezan, A. R. S. (2016). Similarity-based time series prediction. Master’s thesis, Universidade de S˜ao Paulo, S˜ao Carlos.

Parmezan, A. R. S. and Batista, G. E. (2014). Icmc-usp time series prediction repository.

Parmezan, A. R. S. and Batista, G. E. (2015). A study of the use of complexity measures in the similarity search process adopted by knn algorithm for time series prediction. In 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA), pages 45–51. IEEE.

Piazentin, D. R. M. and Rosa, J. L. G. (2014). Motor imagery classification for braincomputer interfaces through a chaotic neural network. In 2014 International Joint Conference on Neural Networks (IJCNN), number 3, pages 4103–4103.

Piazentin, D. R. M. and Rosa, J. L. G. (2015). A simulator for Freeman k-sets in java. In 2015 International Joint Conference on Neural Networks (IJCNN), number 3, pages 1–8.

Rosa, J. L. G. and Piazentin, D. R. M. (2016). A new cognitive filtering approach based on Freeman k3 neural networks. Applied Intelligence, 45(2):363–382.

Stock, J. H. and Watson, M. W. (2005). Implications of dynamic factor models for var analysis. working paper.