Data-Driven Product-Oriented Quality Performance Prediction

This study addresses the characteristics of quality data across different product levels in industrial production and the challenge of cross-level product quality performance prediction. We propose a Bayesian optimization-based dual-layer Long Short- Term Memory (LSTM) network integrated with an attention mechanism. Compared with traditional grid search and random search methods, the Bayesian optimization algorithm demonstrates superior efficiency in identifying globally optimal hyperparameter solutions with minimal evaluation iterations, thereby significantly reducing parameter prediction errors. The proposed attention-enhanced dual-layer LSTM model exhibits notable advantages over conventional LSTM architectures in multiple dimensions: dynamic feature focusing, medium- and long-term dependency modeling, noise resistance, model interpretability, and prediction accuracy. The model was implemented and experimentally validated using Python programming environment, demonstrating its effectiveness in handling complex industrial quality data patterns.