This study analyzes the ability of the recursive dynamics multivaried models constructed through genetic algorithms and the recursive neuronal networks to predict the sign of the IPSA's weekly changes. The data correspond to the period between July 14,1997 and December 9, 2002. The analyzed models were evaluated in 60 series generated by a block-bootstrap process. The results indicate that the network ward would have better forecast ability than the genetic algorithms model and the naive model to predict the sign of the IPSA's changes. The network ward and the genetic algorithms models surpassed, in average, the buy and hold strategy, even when a transaction cost of 0.1 % is considered.
Parisi F., A. (2004). Models of genetic algorithms and neural networks in the prediction of the sign of variation of the IPSA. Estudios De Administración, 11(1), 55–79. https://doi.org/10.5354/0719-0816.2004.56798
