Abstract According to the characteristics of mass incidents in information society, the network geospatial analysis method is firstly proposed in this paper. As the breakthrough point of early warning research, it is a feasible, intuitive and visual early warning method of mass incidents. From the perspective of geospatial analysis, the geographical distribution of the region of event participants has been taken as the breakthrough point in order to analyze the geographical distribution of sensitive population in spatial map. It is required to use the adaptive learning of artificial intelligence to dig the data and extract the rules of online information as well as provide the analysis for network public opinions. Based on the visual geospatial analysis, the early warning of mass incidents can be made in the intuitive visual map, which can dynamically show the online hot contradictions, the distribution and the trends of irritable crowd in the map.