REDUCTION OF PORTFOLIO MODEL FOR EARTHQUAKE RISK ASSESSMENT USING K-MEANS METHOD

Abstract

Evaluating the risk of a large portfolio consisting of a vast number of buildings and structures is computationally intensive, and when considering the correlation between locations based on distances using Monte Carlo simulations, the computational burden becomes extremely large. This study focused on portfolio reduction and proposed a reduction method using the k-means method, as well as a method for determining characteristics after reduction. The proposed method was applied to a portfolio of 346 buildings located in the Kanto region, and its advantages were confirmed. Insights regarding portfolio reduction were also organized through sensitivity analysis.