A risk matrix approach based on clustering algorithm

Abstract

The traditional risk matrix is insufficient in some cases due to the static, non-meticulous classification of probability and severity defined without considering of the context. In order to deal with this issue, this study proposes an extended RMA method (eRIA), by adding probability level and severity level steps in classification based on the basic steps of RMA. With the help of a classic clustering algorithm in data mining, which is K-means, through the iterative learning algorithm, dynamic division of probability and severity is realized. At last, the eRIA is applied to evaluate the traffic block risk and three ranks of risk sources are founded by the risk index. © 2013 Asian Network for Scientific Information.