Comparison of different variable selection methods for partial least squares soft sensor development

Data-driven soft sensors have been widely used in both academic research and industrial applications for predicting hard-to-measure variables or replacing physical sensors to reduce cost. It has been shown that the performance of these data-driven soft sensors can be greatly improved by selecting only the vital variables that strongly affect the primary variables, rather than using all the available process variables. In this work, a comprehensive evaluation of different variable selection methods for soft sensor development is presented. The following seven variable selection methods are considered: stepwise regression (SR), partial least squares with regression coefficients (PLS-BETA), PLS with variable importance in projection (PLS-VIP), uninformative variable elimination with PLS (UVE-PLS), genetic algorithm with PLS (GA-PLS), least absolute shrinkage and selection operator (Lasso), and competitive adaptive reweighted sampling with PLS (CARS-PLS). Their strengths and limitations for soft sensor development are examined using a simulated case study and an industrial case study. Independent tuning datasets are used to optimize each method and to analyze the sensitivity of each method to its tuning parameters. Then independent test datasets are used to compare the prediction performances of PLS soft sensors developed based on different variable selection methods.