The application of information diffusion technique in probabilistic analysis to grassland biological disasters risk

Abstract

Biological disaster risk analysis is a complicated system. The incompleteness (gray areas), the non-clarity (fuzziness) and the uncertainty (randomness) of the data cause many difficulties that must be addressed with the risk assessment. In China, grasshopper and rodent disasters often occur in remote pastoral regions. This causes the monitored data of biological disaster to have a short series and span a large spatial and temporal scale. As available data are small sample in size, the use of risk assessment is often limited. The grassland biological disaster is a complex non-linear system. For the complex non-linear problems, effective conclusion can not be obtained from the accurate probability theory and mathematical statistics theory, but the fuzziness method may be a better method. In this paper, the one-dimension information diffusion technology adopted in evaluating the grassland biological disaster risk for the small statistical sample. The results show that: The information diffusion technology can make up for the information blank caused by the incompleteness of data, can change the single-valued samples into set-valued samples and excavate the internal law contained in the incomplete sample so as to achieve the aim of making full use of the information. It also can be seen that the diffusion results obtained under different starting control points or different interval step sizes have relatively good consistency and continuity. Based on such stability, a biological disaster risk forecast method can be derived, and the risk map using the reciprocals of different transcending probability values to demonstrate the regional differences on the same disaster level was also made by combining with GIS technology. Compared to other mature theories and technologies, the theory and method of fuzzy information optimization processing has its shortcomings especially in the selection of information diffusion function and information diffusion coefficient, and many improvements are needed.