Enhancing Urban Circular Economy Efficiency: Integration of Artificial Neural Networks with Fuzzy Dynamic Network Slack-Based Measure

Research on the urban circular economy (CE) in developing regions often overlooks cross-sectoral interactions, social dimensions, data uncertainty, circularity metrics, and nonlinear trends, underscoring the need for integrated adaptive assessment. To address these gaps, we propose an integrated framework combining a nonlinear autoregressive with exogenous inputs (NARX) neural network and a fuzzy dynamic network slack-based measure (DNSBM) model to evaluate and improve urban CE performance across economic, environmental, and social dimensions in 107 cities of the Yangtze River Economic Belt (YREB) from 2011 to 2023. The results show a steady increase in aggregate efficiency and robustness across α-cut levels, alongside marked regional and stage heterogeneity. Downstream cities perform better because of more effective resource coordination, whereas upstream cities show greater potential for improvement. The main constraint is the social health dimension, reflecting persistent underinvestment in public health. ANN-based slack adjustment enhances efficiency estimation accuracy. Most cities need to reduce redundant inputs, curb pollution emissions, and increase health investment. This study contributes a closed-loop, multidimensional framework that captures temporal dynamics, data uncertainty, and cross-sectoral feedback and supports performance optimization and region-specific sustainability pathways.