Abstract: spatial-temporal big data produced by Location Based Service (LBS) provide an important opportunity for geo-environment simulation, and a novel way for simulation of land use types. The periodicity change of residents activity under the LBS big data has a significant correlation with land use, and show strong land use indication. LBS big data show great application value in land use type simulation. APP alliance user's Location Based Service big data on the Baidu Map platform is used to refer the land use in Guangzhou. The random analysis method in the supervisory classification method is used to simulate the land use type taking the 150*150m grid as the precision unit. The simulation result is total accuracy rate 72.40%, and among which the simulating accuracy rate of villages and towns construction land, industrial land, park green land, residential land, commercial and financial land are 50.39%, 82.65%, 49.53%, 82.66% and 49.56% respectively. The validation result confirms that random analysis method effectively infer land use types based on the high-density geographic behavior big data and demonstrates the potential values of using the LBS big data in urban land use inference, and also reveals the significant linkage between the residents activity spatial pattern and the underlying urban land use. Of course, this requires more case studies in different cities and different regions in the city using different analysis technical means.