Disaster Risk Diagnosis for Traditional Houses Based on a Self-Organizing Map

Abstract

A self-organizing map (SOM) is well-known as a good tool for clustering. Based on such a function to visualize an applied data structure in the form of a feature map, a disaster risk diagnosis for traditional houses in historical regional villages is tried. According to some computer simulations with actual residential data, it is found that overall tendencies in the feature map are developed automatically as if each sample is aligned with its risk. For example, a series of samples, whose risk is almost the same, seems to change its attribute in a trade-off manner to maintain its moderate-risk constantly. In contrast, a series of samples, whose risk increases rapidly, seems to change unilaterally. Then, it is concluded that the proposed method is an effective means to estimate disaster risk appropriately.