New approach for regional water-energy-food nexus security assessment: Enhancing the random forest model with the aquila optimizer algorithm

To uncover the security aspects of the Water-Energy-Food-Nexus (WEFN) and develop innovative methods for evaluating its security, we initially utilize the coefficient of variation-cumulative information contribution rate (COV-CICR) model to select assessment metrics. This approach ensures a well-suited set of indicators for assessing the security of the WEFN. Subsequently, the WEFN security assessment model is developed by utilizing a random forest model that has been fine-tuned with the aquila optimizer. To illustrate this model's application, we use the example of the Hongxinglong branch of China's Beidahuang Farming Group Co., Ltd. The study conducted a detailed analysis of the spatiotemporal variations in the security of the WEFN (Water-Energy-Food Nexus) system across 12 subsidiary farms. Furthermore, it identified key driving factors and elucidated their mechanisms by accounting for data fluctuations. The research included calculating the contribution index of these key driving factors to accurately determine the dominant factors under varying conditions. The results indicate that over time, the WEFN security level in the study trended upward. From 2002–2008, the WEFN security level slowly increased. From 2009–2018, WEFN security remained stable. From 2018–2021, the security level rapidly accelerated. Spatially, the WEFN security in the southern region is lower than that in the northern region, while the central region’s security remained stable. The primary controlling factors for WEFN security also vary at different scales. In comparison to the BP, RF, and PSO-RF models, the AO-RF model demonstrates outstanding advantages in terms of fitting performance, reliability, reasonableness, and stability. This fully supports the applicability of the AO-RF model in WEFN security assessment. The research findings enrich the integrated application of intelligent optimization and machine learning in WEFN studies, introduce a novel analytical model for WEFN security drivers, and expand our understanding of the new mechanisms underlying WEFN security operations.