Risk assessment and comparison of regional natural disasters in China using clustering

Natural disaster that contributes to the economic crisis all over the world has a crucial role in emergency management. The assessment of regional risk to natural disasters is normally studied as a multi-criteria decision making (MCDM) problem in the literature. However little effort was devoted into the comparison of temporary disaster risk of regions. In this paper, a hybrid approach is proposed integrating MCDM and clustering for evaluating and comparing the regional risk to natural disasters. Our two-stage method is applied to thirty-one Chinese regions over the past two consecutive years. In the first stage MCDM is used to prioritize the regions yearly yielding a set of risk vectors over the given period. In the second stage, K-means clustering is applied to divide the regions into a number of clusters characterized by different risk variation patterns. The derived patterns reveal the variation of regions in perspective of natural disaster risk and therefore offer valuable suggestions for disaster risk reduction.