基于改进余弦相似度的证据间定向冲突度量方法

针对传统证据冲突度量方法无法区分证据间冲突程度的差异,并且会在特定情况下失效的问题,提出一种基于改进余弦相似度的证据间定向冲突度量方法。通过将支持系数引入余弦相似度模型,使所提方法具有非对称性,从而能够区分证据间冲突程度的差异,同时还可解决基本余弦相似度模型不适合处理证据中包含非单子集焦元的问题。实验结果验证了所提方法的准确性和有效性。

Directional evidence conflict measurement based on improved cosine similarity

In order to solve the problems that traditional evidence conflict measurement methods can not express the difference of mutual conflict degrees between evidences and work poorly under special situation, a directional evidence conflict measurement based on the improved cosine similarity is proposed. By importing the support coefficient matrix into the cosine similarity model, the proposed model can be asymmetrical, so that the difference of mutual conflict degrees between evidences can be expressed, and the problem that the basic cosine similarity model is not suitable to measure the evidence conflict when the evidences include multi subset focal -elements can be overcame. Simulation results show that the proposed method is effective and accurate.