LASSO-BN for Selection and Optimization of Product Critical Quality Features

The prediction of complex product quality has been extensively studied in the last decades. However, due to the high dimensionality and diversity of quality features, the control optimization of feature parameters remains a significant challenge. For complex products, fault detection techniques are required to accurately predict product quality, and significant influencing quality features must also be identified for their control optimization. In this paper, we propose a novel approach combining the Least Absolute Shrinkage and Selection Operator (LASSO) method with Bayesian Networks (BN) for the detection, identification and control of product quality metrics. Specifically, for complex products with high-dimensional features, the identification of key quality features is achieved initially by the LASSO method to obtain more accurate quality prediction. Subsequently, the optimal production range is determined through the utilization of a Bayesian network to achieve the optimization of product quality. The experimental results show that processing fewer but critical features can not only obtain satisfactory prediction accuracy, but also save computational time. Furthermore, this method offers practical operational guidance for product quality prediction and control in complex product industries.