<p>Understanding the mechanisms driving streamflow changes is crucial for effective water resource management in arid regions. However, basin-scale assessments based on modeling approaches are often constrained by data scarcity and simulation uncertainty. This study explored the relative contributions of climate change and human activities to streamflow changes in the Dahei River Basin in the upstream region of the Yellow River, China, using three different machine learning models (MLMs) to reconstruct natural streamflow from 1980 to 2020 at eight streamflow stations. MLMs are constrained by their reliance on the quality, quantity, and representativeness of the data, which can introduce uncertainties in simulations. To address this limitation, in this study, assessment uncertainties were evaluated on the basis of the differences among MLM simulations. The Mann‒Kendall test indicated that while precipitation remains unchanged, temperature and streamflow show significant increasing and decreasing trends, respectively, except at Xierdaohe station, where streamflow increased. On the basis of monthly data, continuous wavelet transform analysis revealed 1-year periodicity for the three hydrometeorological variables, with the periodicity of streamflow vanishing recently. At seven stations, streamflow decreased after the abrupt change, primarily due to human activities, with relative contributions ranging from 58.7 to 119.8%, likely from groundwater extraction and reservoir operations. Xierdaohe station showed a 64% increase in streamflow, with a human activities impact between 94.6% and 116.3%, possibly due to wastewater treatment&#xa0;plant discharge. The findings of this study are valuable for enhancing our understanding of changes in the water cycle in this arid basin and distinguishing the impacts of climate change from those of various types of human activities.</p>