AI-Empowered Export Resilience Enhancement: An Empirical Study on Guangdong's Trendy Toy Industry Based on Machine Learning and Big Data Analytics

The rapid diffusion of artificial intelligence (AI) and big data technologies is fundamentally reshaping the competitive dynamics of China's creative manufacturing sector. This paper proposes and constructs an AI Application Capability Index (AACI) for the trendy toy industry, integrating five sub-dimensions: AI-driven design adoption, big data marketing analytics, intelligent supply chain management, AI patent intensity, and digital talent density. Leveraging panel data from Guangdong's trendy toy export sector (2014–2024) and customs-level trade statistics covering 110+ destination markets, we apply a multi-model machine learning framework—comprising Random Forest, XGBoost, and LSTM—to predict export resilience and benchmark against traditional OLS regression. XGBoost achieves the highest predictive accuracy \(( {{R}^2 = 0.891,RMSE = 0.079} )\). SHAP-based feature attribution identifies AI design adoption and big data utilization as the two most influential determinants. A mediation analysis further reveals that export growth rate partially mediates the AACI→export resilience relationship (indirect effect ratio: 31.2%). The findings provide both a replicable computational framework and actionable policy guidance for AI-driven export resilience.