Research on quality and safety risk identification of import and export toys based on the WOA-BP model

Within customs risk management, the weak ability to identify quality and safety risks of import and export toys-special products with high safety sensitivity-leads to persistently high recall rates. This study constructs a "data collection-risk classification-risk identification" framework integrating LDA and WOA-BP to address this. First, 2010-2024 toy recall texts from international systems are preprocessed. TF-IDF word frequency analysis and network diagrams reveal key risk words and their correlations. LDA then clusters these words into structured risk factors (types, materials, defects) and five risk event categories, providing model inputs. The WOA-BP model, optimizing BP with the Whale Optimization Algorithm, achieves 95.71% accuracy-5.71% points higher than unoptimized BP and outperforming PSO-BP and GA-BP. Its test set R² reaches 0.997, showing strong fitting. This framework enhances toy risk identification and offers a new method for import-export product safety management.