基于知识元的突发事件风险熵预测模型研究

Emergency risk entropy forecasting model based on knowledge element

突发事件风险预测中受到多因素高维数据和小样本数据信息不完备的约束,无法定量识别和描述事件风险的不确定性.本文根据突发事件已认知的知识要素描述事件的共性本体特征,构建了基于知识元的突发事件风险预测模型.该模型利用状态监测系统的实时数据,根据最佳投影方向对观测样本数据进行降维,将投影特征值隐含的风险信息在风险指标论域内进行扩散,获得突发事件不同风险等级发生的概率,利用风险熵预测突发事件发生的可能性.以2008年长沙突发雨雪冰冻事件为例,对提出的风险预测方法的可行性和有效性进行了验证.研究结果表明,基于知识元的突发事件风险熵能够定量地反映概率事件发生时传递的风险信息,可为应急管理部门科学决策提供依据.

The emergency risk prediction is limited by multi factor, high dimensional data, small samples and incomplete information not fully distinguishing risk of emergency. This paper provided a method of emergency risk prediction based on knowledge element which described the common ontological characteristics of the emergency. By using real-time data of condition monitoring system, this model reduced the dimensionality of the observed data through the optimum projection direction, diffused the risk information contained in the projected characteristic value into the risk index universe, attained the risk probability of emergency risk levels,and forecasted the emergency possibility through risk entropy. The study took the events of freezing rain and snow in Changsha, China, in Early 2008 as an example to prove the feasibility and validity of the risk prediction method. The results showed that the emergency risk entropy based on knowledge element could quantitatively reflect the risk information when the probability event happen. The finding could provide a scientific basis for the emergency management.