A hybrid approach based on neural networks and genetic algorithms to the study of profitability in the Spanish Stock Market

Abstract

This paper studies predictability and profitability of using neural networks (NN) in the Spanish security market. This is carried out through a hybrid approximation which entails evolving a genetic algorithm in order to obtain an optimal NN's architecture. To that end, (NNs) forecasts are transformed into a simple trading strategy, whose profitability is evaluated against a simple buy-and-hold strategy.

Notes

The term backpropagation refers to the process by which dervatives of the network error, with respect to network weights and biases, can be computed.

These selections are common in the NNs literature.

The fitness function depends on the Mean Square Error (MSE) in the following way: fitness = 1/(1 + MSE).

A detailed description of the follwoing training function can be found in Houck et al. (1995). All results are available upon request.

Because of a significant reduction in the number of previous observations (1063 data point) only 60 days ahead forecast have been concentrated on.