Radial Basis Neural Network design using a competitive cooperative coevolutionary multiobjective algorithm

This work presents a new mining model to train Radial Basis Neural Network (RBNN) for short term prediction. Training is performed in two phases. First, weights of the radial basis function (RBF) are trained and in the second phase, a competitive cooperative coevolutionary multiobjective algorithm is used to determine the parameters for each RBF node. The model has been applied to real problems in time series prediction and the obtained results are similar to those obtained by a model representing the state-of-the-art in bioinspired algorithms for time series prediction.