Robust food production systems are fundamental to ensuring food security. In this study, we integrate the conceptual implications of food production resilience and construct a multidimensional evaluation index system that encompasses resistance, resilience, and adaptive capacity. This study utilizes panel data of Xinjiang from 2010 to 2022 and uses integrated methods such as entropy weight method, modified gravity model, social network analysis (SNA), and XGBoost-SHAP model to systematically analyze the resilience level, structural network characteristics, and underlying driving mechanisms of regional food production. The results show that: Production in Xinjiang shows a continuous growth trend and is characterized by spatial heterogeneity with relatively narrow gaps. During the research period, the network connectivity of food production resilience in Xinjiang became increasingly tight. However, it was characterized by a topology with low network density, high clustering, and short average path. Asymmetric features were observed between the input and output regions, accompanied by a decrease in the number of spillovers across the block. The top four influencing factors were per capita cultivated land area (X5), transportation access (X15), agricultural technology progress (X12), and annual average temperature (X1), among which the interaction effect of per capita cultivated land area (X5) and transportation access (X15) was the largest. Insights from these studies can provide valuable references for safeguarding national food security.
