Basin ecological health (BEH) assessment is a critical component of basin ecological restoration and governance. However, existing theories and methods often lead to ineffective management strategies due to their inadequate consideration of uncertainty and the complex driving mechanisms of different factor types on the BEH. To overcome these bottlenecks, this study constructs a research framework that integrates uncertainty assessment methods and identifies the evolution and driving processes of BEH influencing factors, and validates this framework through a case study of the Beiluo River, a typical basin in the semi-arid region of the Loess Plateau that is significantly affected by human activities. This framework establishes a multi-dimensional indicator system from the dual perspectives of natural environmental factors (NEF) and anthropogenic disturbance activities (ADA), and couples a spatial similarity cloud model with the random forest-partial dependence plot (RFPDP) algorithm. It aims to advance traditional multi-indicator assessment into a comprehensive health diagnosis that accounts for uncertainty, while simultaneously enabling visual analysis of the evolutionary processes of BEH influencing factors and the response mechanisms of the BEH to key driving factors. The results show that the BEH of the Beiluo River Basin is generally at a basically healthy level, but the midstream and downstream reaches face more concentrated ecological pressures. This spatial pattern emerges under conditions of a nonlinear increase in anthropogenic disturbance pressure along the river and a relatively stable natural environment. It is against this background that this study, through quantitative analysis, reveals that the dominant influencing factors of the BEH undergo a successional process. That is, after initially crossing a critical point under high-intensity anthropogenic pressure, the BEH's resilience is weakened, and it subsequently becomes highly sensitive to previously bufferable natural environmental fluctuations. This succession exhibits spatial heterogeneity due to geomorphology and socioeconomic differences. Specifically: physical disturbances are dominant in the upstream reaches; the midstream reaches are subject to compound multi-source impacts, such as water quality and flow variation; and pressures on flood control infrastructure are highlighted in the downstream reaches. Furthermore, this study reveals that the response of the BEH to most influencing factors follows a nonlinear process with distinct thresholds, providing a targeted reference for precise regulation. Based on the identified threshold effects, adaptive management measures are proposed. The core of upstream management is to stabilize the health score of the riparian zone disturbance indicator within the high-efficiency sensitive interval of 79-89, as the BEH shows a more significant improvement within this range, thus maximizing ecological investment benefits. The midstream reaches require coordinated regulation to ensure the water quality score breaks through the activation threshold of 40 while maintaining the ecological flow satisfaction rate within the effective interval of 45-60 %, thereby promoting the continuous improvement of the BEH. The downstream reaches should focus on increasing the flood control compliance rate to above the critical threshold of 96.5 % to trigger a positive response in their BEH, thereby ensuring the ecological security pattern for their agricultural and residential areas.