Neural network ensemble modeling for nosiheptide fermentation process based on partial least squares regression

Abstract

Nosiheptide fermentation product concentration model based on neural network ensemble is presented. Data for building the model is re-sampled from the original training data using Bagging approach. For each pair of training data an individual Elman neural network is trained. Then outputs of the individual neural network are then combined to form the overall output of the neural network ensemble through the weighted average method and the combining weights are determined by partial least squares regression. The model built on neural network ensemble is compared to a single neural network model, and the results show that it has high accuracy and generalization ability.