基于CD-型贴近度的云模型相似性度量方法

The Similarity Measurement Method of Cloud Models Based on CD-Type Closeness

摘要: 在云模型的实际应用中，云模型相似性度量是极为关键的环节。针对现有云模型相似性度量方法中存在的区分度欠佳、结果不稳定等问题，本文以正态云的扩展模型三角云为研究对象，在充分考虑期望曲线及最大边界曲线基础上，将其视作三角模糊数，通过计算三角模糊数的CD-型贴近度来度量云模型的相似性，从而提出一种求两云模型相似度的CDTCM综合计算方法。由仿真实验可知，提出的方法具有一定的区分度；在Synthetic Control Chart Dataset数据集上的分类对比实验表明，该方法的分类精度优于ECM、MCM、LICM、CFSM、EMTCM等传统方法；对于UCR数据库中的10个数据集表现出良好的分类效果，验证了该方法具有一定的可行性及有效性。

Abstract: In the practical application of cloud models, similarity measurement is a critical component. Aiming at the issues of poor discrimination and unstable results in existing cloud model similarity meas-urement methods, this study focuses on the triangular cloud, an extended model of the normal cloud. By fully considering the expected curve and maximum boundary curve, the triangular cloud is treated as a triangular fuzzy number. The CD-type closeness degree of triangular fuzzy numbers is calculated to measure the similarity of cloud models, and a comprehensive calculation method named CDTCM for evaluating the similarity between two cloud models is proposed. Simulation ex-periments show that the proposed method has a certain degree of discrimination. Classification comparison experiments on the Synthetic Control Chart Dataset indicate that the classification ac-curacy of this method outperforms traditional methods, such as ECM, MCM, LICM, CFSM, and EMTCM. Additionally, it demonstrates good classification performance on 10 datasets from the UCR database, verifying that the method is stable and demonstrates certain feasibility and effectiveness.