L. Anastasakis and N. Mort (UK)
Neural networks, ARMA, forecasting and prediction, fi nance
Financial systems are characterized by complex ob jects, noisy data and nonlinearities. Conventional methods as well as artificial intelligence techniques have been ap plied to beat the weak form of efficient market hypothesis. In this paper the traditional Box-Jenkins approach as well as an MLP neural network are found to perform better than the random walk model. Neural networks is proved to extract the nonlinearities hidden in the data and exhibit significant profit in comparison to ARMA models and the naive buy and hold trading strategy. A neural network modelling approach to find the optimum number of hid den units is also presented.
Important Links:
Go Back
