The spatial distribution patterns of land cover greatly influence the ecological balance of the Loess Plateau. Understanding the bio-physical drivers of land cover change is important for ecological restoration in the context of climate change. However, in the analysis of the drivers of land cover change in the Loess Plateau, the role of bio-physical drivers has not been quantitatively evaluated. Using remote sensing data, machine learning, and statistical methods, this study analyzed the spatial and temporal patterns of land cover from 2001 to 2018 in the Loess Plateau of China. We used a random forest (RF) model to quantify the importance of bio-physical drivers of land cover. Our results demonstrated that the RF model has good performance and high reliability (model accuracy score > 0.8). Our simulation experiment revealed that evapotranspiration was the most important driver (importance score, IS >0.2), temperature and precipitation had regional heterogeneity, and slope was the least important (IS <0.05). We suggest that evapotranspiration can be regulated by properly allocating the type of land cover, so as to rationally allocate water resources on the Loess Plateau. This study provides a new foundation for quantitatively evaluating the drivers of land cover change and regulating the distribution of water resources on the Loess Plateau, China.