In this paper, to naturally fill the gap in incomplete data, a new algorithm is proposed for estimating the risk of natural disasters based on the information diffusion theory and the equation of the vibrating string. Two experiments are performed with small samples to investigate its effectiveness. Furthermore, to demonstrate the practicality of the new algorithm, it is applied to study the relationship between epicentral intensity and earthquake magnitude, with strong-motion earthquake observations measured in Yunnan Province in China. The regression model, the back-propagation neural network and the conventional information diffusion model are also involved for comparison. All results show that the new algorithm, which can unravel fuzzy information in incomplete data, is better than the main existing methods for risk estimation.