An automated investment strategy using artificial neural networks and econometric predictors

Information systems with the objective to make forecasts for financial time series and negotiate from these are subject to various risks, because the stock market is influenced by diffe rent sources continuously. The study of quantitative finance addresses methods for treating problems such as these, a fact which occurs mainly through the use of computational intelligence. This paper presents an automated strategy (investor robot) that combines predictions made by artificial neural networks and econometric predictors in a second neural network, this acts like a ensemble. The predictions are used to generate purchase or sell signals through a negotiation model built into the algorithm. The experiments were conducted with real series of three assets with high liquidity, a commodity and a market index. The financial results are compared against the individual application of each predictor and also the classical market techniques.