Based on the daily runoff data of the hydrological station in the Lake Poyang Basin (from 1960 to 2013) and the daily meteorological data of 14 national meteorological stations, this paper applied the long-short-term Memory network (LSTM) to simulate the runoff dynamics in the basin and attribute the contributions of human activities and climate change on eco-flow variations using eco-flow indices. The results show that:(1) The contribution rate of summer runoff in the five stations reached more than 40%, and the ecological surplus was mainly in summer. The contribution rate of summer runoff in S0 was higher than that in S1 and S2, which may be due to the water storage project. (2) Annually, the changes in the eco-flow indices and precipitation are relatively consistent in the entire period (with correlation coefficients above 0.7), however, the correlation between the eco-flow indices and precipitation is lower at the seasonal scale. (3) Inter-annually, climate change dominant the variation of ecological surplus (60%-85%) in the basin except for the Ganjiang River (26%). Except for the Raohe River (25%) and Fuhe River (52%), the contribution of climate change on the variations of eco-flow indices are 25% in Raohe River, 52% in Fuhe River, 95% in Ganjiang River, 98% in Xinjiang River, and 99% in Xiushui River, respectively. (4) At seasonal scale, climate change causes the reduction of eco-deficit and the reduction is strongest in spring and winter. Whereas the eco-surplus increased in most cases. Human activities cause the increase of eco-deficit, mostly in autumn and winter, and it causes a decrease in eco-surplus in winter. Our research proposed a new method for attributing the impact of climate change and human activities on hydrological changes. The results should provide theoretical support for water resources management in the Lake Poyang Basin.