自然灾害风险分析的信息矩阵方法

从致灾因子危险性分析，到伤亡和损失的可能性研究，自然灾害风险分析由4个环节组成，即致灾因子、承灾体输入、承灾体输出和社会系统。在大多数据情况下，风险事件的不确定性，主要来自于致灾因子时间、地点、强度的不确定。因此，自然灾害风险分析主要涉及两类模式识别：致灾因子概率分布识别和承灾体系统输入-输出关系识别。由于概率分布和输入-输出关系在数学上均可用函数表达，所以，自然灾害风险分析涉及的两类模式识别，均是函数关系的识别。给出了用信息矩阵识别这些函数关系的方法。为演示该方法的优点，用它处理了在我国得到的地震数据，识别了震中烈度与震级的关系。结果表明，得到的结果比线性回归、基于正态分布的模糊推理和人工神经元网络等的结果好。

Information matrix method for risk analysis of natural disaster

From analyzing dangers of a hazard factor to studying possibilities of casualty and loss,there are 4 steps in risk analysis of natural disaster,i.e.,a hazard factor,the input to a hazard-affected body,the output from a hazard-affected body,and the society system.In the most cases,the uncertainties of risk events are due to that we do not know when,where and how a hazard factor will occur.Therefore,risk analysis of natural disaster mainly relates to two kinds of patterns to be recognized: the probability distributions of hazard factors and relationships between inputs and outputs in systems of hazard-affected bodies.The probability distributions and the input-output relationships can be mathematically represented by functions;therefore these recognitions are the function recognition.In this paper,a novel approach, called information matrix,was introduced to recognize these function relationships.To display the advantage of the model,in this paper,it was used to recognize the relationship between magnitude and epicentral intensity with the earthquake data recorded in China.The result from the information matrix model is better than ones from linear regression model,fuzzy inference based on normal assumption,and artificial neural networks.