Research And Application Of Prototype Clustering Algorithm Based On Three-way Decision

Prototype clustering algorithms generally believe that the clustering structure can be described by a set of prototypes.This type of algorithm is simple and efficient,and is very common in real-world clustering tasks.Among them,the Gaussian mixture clustering algorithm and the K-means algorithm are more commonly used.The Gaussian mixture clustering algorithm assigns cluster members based on probability,which is easy to cause the algorithm to have a greater risk of misjudgment.In this thesis,the three-way decision theory is combined to conduct research,and a Gaussian mixture clustering algorithm based on the three-way decision is proposed.The experiment further shows that the prototype clustering algorithm based on the three-way decision has better clustering performance.Specific work includes:(1)Propose a Gaussian mixture clustering algorithm(T-GMM)based on three-way decision.This algorithm takes into account that the Gaussian mixture model often has unclear membership in clustering,that is,the probability that the sample points belong to multiple categories is close,so the three-way decision ideas are integrated into the Gaussian mixture model to improve the performance of the original algorithm.The new algorithm reduces the risk of directly making a decision that the object belongs to or does not belong to a certain class,and effectively reduces the cost of misjudgment.The experiment further shows that the algorithm proposed in this thesis not only inherits the characteristics of the Gaussian mixture clustering algorithm,but also has good clustering performance,and shows excelent clustering effect for non-spherical data clusters.(2)Apply the new algorithm to medical data diagnosis.This thesis selects prostate tumor biochemical data for clustering application analysis.After cleaning and nondimensionalizing the source data,the T-GMM algorithm and the K-means algorithm based on three-way decision are respectively applied and compared and analyzed.The results show that the new algorithm combined with the three-way decision theory is effective in researching and analyzing the severity of patients' prostate tumors.The new algorithm identifies the characteristics of data in advance and makes good decisions.It has certain reference significance in practical application,and can also provide decision-making basis for the research of medical diagnosis.