Abstract: To study the spatial distribution of population density in illegal buildings in Shenzhen,the authors used GIS technologies such as spatial autocorrelation analysis method and regression analysis method. Spatial autocorrelation analysis method was used to study the spatial distribution pattern of the population density in illegal buildings in Shenzhen. Both monocentric population density models and multi-centric population density models were fitted by regression analysis method. The study results show that the spatial clustering pattern of the population density is existent according to the global Moran's I index, and the area joining the Futian, Luohu and Longhua district is the hot spot area according to the Local Moran's I and Getis-Ord Gi? index. The fitting results show that,among all the population density models,the quadratic exponential model can best describe population density distribution in illegal buildings of Shenzhen, while the overall fitting results of multi -centric models are at low level, which shows that the multiple centers fail to form the population in illegal buildings. The quadratic exponential model can not only obtain the best result from statistical point of view but also verify that it fits the crator theory well with rent data of illegal buildings.