To address the challenges of multi-dimensional evaluation in agricultural sustainable development, this study integrates the composite evaluation method with multi-level factor analysis to construct an integrated model comprising “indicator dimension reduction-weight assignment-spatial validation.” First, principal component analysis (PCA) and entropy methods are combined, and after passing the Spearman consistency test (ρ < 0.05), the fuzzy Borda model is used to synthesize the evaluation results. Subsequently, factor analysis is used for dimension reduction. The KMO value of 0.892 (Bartlett's test P=0.000) supports the extraction of three principal components. After rotation, the cumulative contribution rate reaches 85.766%. Four indicators with loadings <0.4 are excluded, ultimately establishing 18 core indicators across three categories: resource environment (9 indicators), production economy (9 indicators), and population and society (4 indicators). Empirical analysis of data from Region A from 2020 to 2024 indicates that resource pressure has intensified, with per capita arable land (C1) decreasing by 14.3% to 0.12 hm²/person. However, ecological governance has achieved significant results, with the proportion of soil erosion (C6) decreasing by 20.0%. The economic dimension dominated the comprehensive evaluation (AHP weight of 62.8%), with agricultural total output value (C9) having the highest weight of 0.118. Regional evaluation results showed that all 20 regions scored an average of 13.26 (Grade II, good), but the range was as high as 11.95 points (Region k scored 17.19 points while Region l scored 5.24 points), indicating significant spatial differentiation. The Moran's I scatter plot reveals the expansion of high-value clusters (HH) from 25% to 40% between 2020 and 2024, while low-value zones (LL) shrink, reflecting policy coordination driving regional balanced development.
