Would Evolutionary Computation Help for Designs of Artificial Neural Nets in Financial Applications?

Since the pioneering work by White (1988), the application of artificial neural networks (ANNs) to finance has enjoyed an exponential growth in research and publications. The evidence accumulated over the last decade indicates that the success of the financial application of an ANN depends crucially on its design. The last few years have seen a series of financial applications of evolutionary ANNs (EANNs). Margarita (1991) applies a genetic search to the weights of a recurrent network for the trading of the FIAT shares in the Milan Stock Exchange. In Dorsey, Johnson and Mayer (1995), the GA is found to perform well when optimizing neural networks (NNs). Sexton, Johnson and Dorsey (1995) also find the GA-optimized NN to outperform the back-propagated NN (BPNN) when testing out-of-sample, thereby addressing the problem of overfitting. Harrald and Kamstra (1998) use evolutionary programming to replace the more familiar back-propagation method to fine tune the connection weights of feedforward nets for forecasting volatility. White (1998) shows that a Genetic Adaptive Neural Network (GANN) is able to approximate, to a high degree of accuracy, the complex, nonlinear option-pricing function used to produce simulated option prices. While these studies clearly evidence the promising feature of using evolutionary computation in the design of ANNs, in Yao's categorization, they are all concerned with the lowest level of evolution, namely, connection weights. Other dimensions of the design of an artificial neural net, such as the number of hidden layers, number of hidden nodes, inputs, and transfer functions have not been tackled. Therefore, the purpose of this paper is to extend the current financial applications of EANNs to a higher level of evolution and to evaluate their relevance.