A Regularized Mixture of Linear Experts for Quality Prediction in Multimode and Multiphase Industrial Processes

This paper proposes the use of a regularized mixture linear experts (MoLE) for predictive modeling in multimode-multiphase industrial processes. For this purpose, different MoLE were evaluated, namely, through elastic net (EN), Lasso, and ridge regression (RR) penalties. Their performances compared when trained with numbers samples, comparison to other nonlinear models. The models evaluated on real multiphase polymerization process data. Lasso penalty provided best performance among all regularizers MoLE, even small number samples.