Novel fuzzy similarity measures and their applications in pattern recognition and clustering analysis

Fuzzy similarity measures are utilized to match two or more records and are essential to deal with data classification and pattern-matching problems. We have noticed that the existing studies on similarity measures in the classical fuzzy framework have certain issues, for example, inappropriate identification of structured linguistic variables, inappropriate classification results, etc. In this paper, we propose three new fuzzy similarity measures based on continuous functions and realize their advantages in connection with their application to pattern recognition and cluster analysis. The validity of clusters is also identified using the concept of cluster validity index. The experimental results demonstrate that the proposed similarity measures show higher accuracy in the identification of structured linguistic variables and a higher degree of confidence in the classification of unknown patterns. Several application examples with artificial and real data are utilized to demonstrate the credibility and advantages of the proposed similarity measures.