Inefficient utilization puts the productivity of cultivated land in a low development state. The key challenge for the efficient utilization of cultivated land is to clarify how various factors affect the spatial differentiation pattern of cultivated land productivity (CLP), to improve food production. However, evaluating the impact of the intensity and direction of CLP on a large-scale is a difficulty and there is a gap in knowledge. In this study, we used net primary productivity (NPP) to calculate the productivity of cultivated land and reveal its spatial differentiation. Also, this study examined the spatio-temporal heterogeneity of CLP and determined the effect intensity and the direction of effect of various factors on productivity, using the Songhua River basin (SRB) in China as a research case. We used genetic algorithms to modify and improve a neural network model of factor dimensionality reduction, combined with path analysis, cluster analysis, and regression analysis, to identify the main factors impacting CLP, synergies between these factors, and effect intensity and direction. The results showed that: (1) the area of cultivated land in SRB decreased, but the NPP of cultivated land area increased, during 2000-2020; (2) spatially, NPP was relatively low in the middle of the basin and gradually increased towards the periphery; (3) The main positive factors were the normalized difference vegetation index (NDVI), slope, precipitation, evapotranspiration, and total nitrogen, while the main negative factors were temperature, ratio vegetation index, and total phosphorus. Individual principal factors and the synergy between these factors gave CLP different temporal and spatial heterogeneity. Collaborative management of the threshold range of various influencing factors would improve the CLP. This novel information on