Time-series prediction modelling based on an efficient self-organization learning neural network

Abstract

For solving some process control engineering problems which can be treated as a time-series, a fast and accurate self-organization learning strategy is proposed based on the significance evaluation of hidden neurons with respect to the network output. This approach is introduced to optimize the architecture and parameters of span-lateral inhibition neural network (S-LINN) simultaneously. The insignificant neuron(s) will be pruned automated step by step based on the determination of significance index. The proposed self-organizing approach has been tested on one time-series prediction benchmark problem. Simulation results demonstrate that the proposed method has good exploration and exploitation capabilities in terms of searching the optimal structure and parameters for S-LINN.