A SNCCDBAGG-Based NN Ensemble Approach for Quality Prediction in Injection Molding Process

Abstract

This paper presents a SNCCDBAGG-based neural network (NN) ensemble approach for quality prediction in injection molding process. Bagging is used to create NNs for the ensemble by independently training these NNs on different training sets. Negative correlation learning via correlation-corrected data (NCCD) is used to achieve negative correlation of each network's error against errors for the rest of the ensemble by training transformed target data for NN in the ensemble as the desired network output for some epochs. A selection-based strategy is proposed to improve generalization ability when combining Bagging and NCCD. Experimental results show its good performance on quality predicting in injection molding process compared with single NN predictor and NCCD predictor.