<p>Balancing urban development and ecological conservation through coordinated land-use structure remains a core challenge in land-use planning. This study aims to identify the optimal proportion ranges of land-use types under the dual objectives of ecological protection and urban development by developing an analytical framework integrating Extreme Gradient Boosting (XGBoost), Shapley Additive exPlanations (SHAP), and Accumulated Local Effects (ALE). Taking the Central Plains Urban Agglomeration as a case study, four ESs, including water retention, soil conservation, windbreak and sand fixation, and carbon storage, were quantified using the InVEST and RWEQ models. Meanwhile, four urban development indicators, including population density, economic density, the share of tertiary industry in GDP, and urbanization rate, were derived from statistical data. An XGBoost model was then used to establish relationships between land-use type proportions and these indicators, which were interpreted using SHAP and ALE to identify contribution patterns, local effects, and threshold ranges, thereby determining group-optimal and group-suboptimal ranges for land-use type proportions. The results showed that the group-optimal proportion ranges of land-use types in the Central Plains Urban Agglomeration were 32.17%–37.91% for cropland, 31.40%–46.3% and 48.43%–52.69% for forest, 0.31%–0.78% and 1.27%–1.45% for shrubland, 3.1%–4.23% for grassland, 3.08%–5.89% for waterbody, and 19.01%–30.73% for impervious surface. In contrast, bare land shows a very low proportion and predominantly negative effects, indicating that its proportion should be maintained at a relatively low level. To accommodate urban heterogeneity and planning constraints, we further proposed group-suboptimal ranges for land-use type proportions to guide flexible adjustments under the constraint that proportions sum to 100%. The framework offers quantitative, data-driven guidance for defining land-use type proportion ranges in regional planning.</p>