Research on complex product quality prediction method based on FA-lightGBM

The data generated during the manufacturing of complex products exhibits high dimensionality, strong inter-feature correlations, and data imbalance. These characteristics pose significant challenges to the accurate prediction of product qua lity. To address this issue, a complex product quality prediction model employing Factor Analysis and Light Gradient Boosting Machine (FA-lightGBM) was proposed. Initially, the mutual information method was utilized to select the optimal feature subset. Subsequently, a concise set of factor features encompassing all quality information was constructed through factor analysis. Building on this, a quality prediction model was developed using lightGBM, establishing a mapping relationship between manufacturing data features and quality characteristic values. Experimental validation confirmed that the proposed quality prediction model exhibited higher prediction accuracy compared to Lasso regression and eXtreme Gradient Boosting (XGBoost) models.