To alleviate the limitations of statistical based methods of forecasting of exchange rates, soft and evolutionary computing based techniques have been introduced in the literature. To further the research in this direction this paper proposes a simple but promising hybrid prediction model by suitably combining an adaptive autoregressive moving average (ARMA) architecture and differential evolution (DE) based training of its feed-forward and feed-back parameters. Simple statistical features are extracted for each exchange rate using a sliding window of past data and are employed as input to the prediction model for training its internal coefficients using DE optimization strategy. The prediction efficiency is validated using past exchange rates not used for training purpose. Simulation results using real life data are presented for three different exchange rates for one–fifteen months’ ahead predictions. The results of the developed model are compared with other four competitive methods such as ARMA-particle swarm optimization (PSO), ARMA-cat swarm optimization (CSO), ARMA-bacterial foraging optimization (BFO) and ARMA-forward backward least mean square (FBLMS). The derivative based ARMA-FBLMS forecasting model exhibits worst prediction performance of the exchange rates. Comparisons of different performance measures including the training time of the all three evolutionary computing based models demonstrate that the proposed ARMA-DE exchange rate prediction model possesses superior short and long range prediction potentiality compared to others

This paper develops and assesses the performance of a hybrid prediction model using a radial basis function neural network and non-dominated sorting multiobjective genetic algorithm-II (NSGA-II) for various stock market forecasts. The proposed technique simultaneously optimizes two mutually conflicting objectives: the structure (the number of centers in the hidden layer) and the output mean square error (MSE) of the model. The best compromised non-dominated solution-based model was determined from the optimal Pareto front using fuzzy set theory. The performances of this model were evaluated in terms of four different measures using Standard and Poor 500 (S&P500) and Dow Jones Industrial Average (DJIA) stock data. The results of the simulation of the new model demonstrate a prediction performance superior to that of the conventional radial basis function (RBF)-based forecasting model in terms of the mean average percentage error (MAPE), directional accuracy (DA), Thelis’ U and average relative variance (ARV) values