Abstract:

A disaster loss assessment method based on the Dempster-Shafer theory of evidence for storm surge is proposed in this paper. Because of uncertainty of storm surge in the disaster process, we select representative indicators (maximum height of storm surge, significant wave height, disaster prevention and reduction ability) to produce several pieces of evidence. The weight of evidence is calculated by using correlation between indicator and direct economic loss of storm surge disaster. A modified Murphy method is used to fuse evidence from different sources to judge the disaster loss level. An example is used to show that the classification accuracy of the method used in this paper can reach 93.1%, which is better than some existing methods, such as the Naive Bayes, Support Vector Machine, Neural Network, and Decision Tree methods. In addition, the method is simple in computation, and the results of disaster loss assessment can be more detailed with increasing training samples.

Key words: storm surge disaster, loss assessment, evidence theory, disaster loss grade, Murphy method