Aiming at the difficult problems of sparse observational sample data and inadequate collection of information in the process of risk analysis and disaster evaluation on earthquake, event (especially on the devastating earthquake disaster), as well as aiming at the shortcoming of current common information diffusion model which is difficult to describe the uneven structure of actual observational sample data, an uneven information diffusion idea was introduced, the corresponding theory deduce and algorithm improvement was made and an improved “ellipse style” information diffusion assessment technique model was established in this paper. It can efficiently and reasonably diffuse the information of sparse observational data and uncompleted sample into a full fuzzy set. As a simulation application case, this technique was used to evaluate the epicenter intensity and destroy area with the collected earthquake data in Sichuan and Yunnan provinces from 1988 to 2001. By comparing the evaluation results with that of normal information diffusion method, it indicated that the uneven information diffusion method and corresponding “ellipse style” evaluation model can truly model the process of information diffusion and its evaluation result was close to the actual cases.