Inner Mongolia is an important green ecological barrier in northern China, and grassland degradation in this region is of great concern. Fractional vegetation cover (FVC) is one of the most direct indicators of grassland ecological status. At present, it is still a challenge to build an accurate FVC estimation model for dynamic analysis of a large region over a long period of time. In this study, we used a large ground survey dataset, MODIS remote sensing data, and meteorological data from 2000 to 2020, and applied the random forest model for FVC partition modeling and prediction. The Sen+Mann-Kendall trend analysis method and Hurst index method were used to analyze the spatio-temporal changes in the FVC and its future trends. The main results were as follows: 1) The precision of each partition of the random forest model was better than that of the whole region, which effectively reduced the impact of spatial heterogeneity. 2) In Inner Mongolia, the grassland FVC generally showed a spatial pattern of being high in the east and low in the west, with obvious spatial differences. 3) In the past 21 years, the FVC of grassland in Inner Mongolia showed a fluctuating upward trend overall, the area of increased FVC was larger than the area of decreased FVC, and the magnitude of the extremely significant increase and significant increase was greater than that of the extremely significant decrease and significant decrease. 4) In the future, the grassland FVC in Inner Mongolia will generally improve. The area of FVC growth is larger than the area of FVC decrease, and the area with extremely significant growth and significant growth accounts for a high proportion (25.9%) of the total area. It is predicted that vegetation growth will develop well in the future.