A Spiking Neural Network for Financial Prediction

In this paper a Polychronous Spiking Network was applied to financial time series prediction with the aim of exploiting the inherent temporal capabilities of the spiking neural model. The performance of this network was benchmarked against two “traditional”, rate-encoded, neural networks; a Multi-Layer Perceptron network and a Functional Link Neural Network. Three non-stationary datasets were used to test these simulations: IBM stock data; US/Euro exchange rate data, and the price of Brent crude oil. The experiments demonstrated favourable prediction results for the spiking neural network in terms of Annualised Return, for both 1-Step and 5-Step ahead predictions. These results were also supported by other relevant metrics such as Maximum Drawdown, Signal-To-Noise ratio, and Normalised Mean Square Error. The results suggest that the inherent temporal characteristics of the polychronous spiking network make it a more suited architecture than traditional neural networks for use in non-stationary financial data prediction environments