Forex Prediction Based on SVR Optimized by Artificial Fish Swarm Algorithm

Abstract

Taking the radial basis function as a kernel function, a prediction model is developed based on the support vector regression machine (SVR). The optimization of the model parameters, including penalty factor and kernel function variance, is realized by the artificial fish swarm algorithm. The model is used to predict nine foreign exchange rate data with updating and rolling. At the same time, simulating by the cross validation, genetic algorithm, particle swarm optimization algorithm and then evaluating the results from the total error (TE), relative error (RE), absolute root mean square error (ARMSE) and correct trend rate (CTR) comprehensively, the comparison shows that the errors of the model based on SVR optimized by artificial fish swarm algorithm are all minimum and CTR are maximum. In the end, in order to improve the convergence speed and precision further, the self-adaption artificial fish swarm algorithm is presented which is joined the attenuation factor and based on the average distance visual. The result is ideal. Therefore, SVR optimized by the improved artificial fish swarm algorithm can be effectively used in forex prediction.