Research On Risk Matrix Design Method From The Perspective Of Clustering

Risk is a triad consisting of specific scenarios and the likelihood of those scenarios occurring and the corresponding consequences.The risk matrix has become the most widely used risk assessment tool in the field of risk management because of its simplicity and straightforwardness.It is widely used in areas including terrorism risk analysis,asset risk management,food safety risk assessment,healthcare,and rapid risk assessment of public health events.Risk matrices typically measure the magnitude of risk by the two main dimensions corresponding to the risk,i.e.,likelihood and consequence of the event,with each set of dimensions having a color matching the cell to reflect the risk level.When using the risk matrix,only the two input dimensions of the risk matrix need to be evaluated accordingly,and the risk matrix can match the corresponding risk level,which is usually represented by a different color in the risk matrix.Although risk matrices have been widely used in many fields,the lack of standardized design mechanism,inconsistency of risk measures and the logical defects of risk matrices themselves have caused scholars to criticize their use.This paper addresses the most important issue of the lack of risk matrix design criteria and conducts an in-depth study to systematically and comprehensively analyze the nature of risk matrices as follows:(1)A general set of risk matrix design methods is proposed from the perspective of clustering analysis.This paper illustrates the feasibility of the clustering design risk matrix from two aspects:the design nature of the risk matrix and the purpose of using the risk matrix,and discusses the reasons and challenges of using the traditional systematic clustering method to design the risk matrix,combines the inherent characteristics of the risk matrix and the systematic clustering idea,proposes a new rating principle and the corresponding algorithm,and demonstrates the use of this design method through an arithmetic example to verify the improved The practicality of the improved systematic clustering method for designing risk matrices is verified.The research results show that the design of risk matrix is similar in nature to clustering,and the clustering approach of systematic clustering is more suitable for the design of risk matrix than dynamic clustering.The reason for not being able to design the risk matrix with the systematic clustering method is that it is difficult to select similarity indicators suitable for the characteristics of the risk matrix according to the clustering principle of the systematic clustering method.The improved systematic clustering rule combines the characteristics of the risk matrix and the idea of systematic clustering,uses the logical comparison of cells,is able to embed the special distribution information of the risk value points of the cells of the risk matrix in the clustering similarity relationship,and establishes special clustering design rules for the risk matrix.The risk matrix designed based on this method is more practical and universal,and can overcome the defects of low resolution of risk levels in some methods of designing risk matrices.Second,the risk matrix can be designed with the corresponding number of risk levels according to the needs of risk assessors and practices.(2)Based on the characteristics of the risk matrix cells,this paper proposes a new algorithm for cluster design based on numerical simulation comparison,which reduces the complexity of cluster calculation.The results show that when clustering is performed by the systematic clustering method,this improved definition of clustering similarity proposed in this paper is more applicable if the case of infinite points in the initial sample is encountered and a similarity index that can represent the characteristics of the sample cannot be found.Its advantage over inter-class distance,which is a measure of sample similarity in traditional systematic clustering,is that it is able to obtain rating results directly from the initial probability comparison matrix,without the need for a step-by-step clustering matrix as in traditional systematic clustering(3)The essence of many different risk matrix design methods is systematically analyzed,focusing on the similarities and differences between the improved systematic clustering method and the sequential update method,and the advantages and disadvantages of the two methods are compared in terms of qualitative and quantitative evaluation indexes of the risk matrix.The research results show that the risk matrix design methods proposed by scholars essentially include the idea of clustering,and the most common feature of the improved systematic clustering method and the sequential updating method is that they can embed the risk value distribution information of the risk cells in the rating process,and both methods can design risk matrices of corresponding levels according to the requirements of risk assessors.However,the improved systematic clustering method is more suitable for the purpose of risk matrix design in terms of two qualitative indicators,namely,wholeness and uniformity,and slightly better in terms of three quantitative indicators,which are correct point-pair rate,distinguishable point-pair rate and correct decision rate.There is no clear standard in academia on how to standardize the design of risk matrices.This paper systematically analyzes the nature of risk matrix,combines clustering with risk matrix design,and proposes a universal risk matrix design method,which is of great significance for improving the accuracy and reliability of risk assessment and expanding the application of risk matrix,and also provides strong theoretical support for the application of risk matrix in practice.