A Dendritic Neuron Model with Nonlinearity Validation on Istanbul Stock and Taiwan Futures Exchange Indexes Prediction

Time series prediction problems are very important not only because its practical values but also owing to its theoretical relations with other fundamental scientific issues. Many forecasting models have been proposed to deal with this problem, among which those sophisticated models are too complex and time-consuming for simple, especially linear problems. Thus, in this paper we propose a flexible methodology by combining a dendritic neuron model with statistical test, where the former is used to perform the forecasting for time series and the latter is devoted to validating the nonlinearity of the data. Experimental results based on Istanbul stock and Taiwan futures exchange indexes verify the superiority of the proposed method.