Fuzzy set theory (FST) has become a dominant model to deal with imprecision, uncertainty, vagueness, and ambiguity arising from many real-world applications. There are a couple of extensions and generalizations of FST, and an interval-valued FST is one of the important generalizations where the membership values are expressed in terms of closed intervals in [0, 1]. Finding the correlation coefficient of different FSTs can be an interesting research problem, as it has many real applications, from social science to medical science. This article first proposes a new extension of fuzzy set in terms of an interval-valued membership function and an interval-valued reference function and then proposes a new method of finding the correlation coefficient using the statistical parameters, like covariance and variance of the proposed fuzzy sets. The proposed formulae can be applied to both discrete and continuous sets of the universe of discourse. It has also been found that the correlation coefficient computed by the proposed formula lies in the closed interval of −1 and 1, which establishes that the correlation coefficient of the proposed fuzzy sets not only gives the strength of the relationship but also tells whether they are positively or negatively correlated. Furthermore, two practical applications along with numerical examples are discussed in detail. The result convincingly establishes the efficacy of the proposed formula.
The most challenging aspect of a pattern recognition problem is to identify a pattern with the help of its positive and negative features. The bipolar fuzzy sets (BFSs) are the mathematical tools that can express a pattern with the help of positive and negative membership functions, hereby making it possible to be expressed more conveniently than others. In this article, a real-valued function on the set of BFSs over a universe of discourse is proposed that fulfills the axioms of being a metric measure on the set of BFSs. The proposed metric can be formulated for both the discrete and the continuous universes of discourse. The efficacy of the proposed metric is validated through several important mathematical properties. Further, a bipolar fuzzy clustering algorithm for sentiment analysis is discussed in detail to demonstrate the usefulness of the proposed metric.
Complex interval-valued fuzzy sets are powerful tools for representing uncertainty and periodicity that occurred in many real-life problems. They not only take periodicity semantics and uncertainty together into account but also describes the information using interval valued membership grades, which gives the expert more freedom to solve complex real-life problems reasonably. Complex interval-valued fuzzy sets have been successfully employed many times in medical diagnosis and pattern recognition problems. In this article, two novel methods for computing the correlation coefficient and weighted correlation coefficient of complex interval-valued fuzzy sets are proposed. The methods employed fuzzy statistical parameters such as mean, variance, and covariance of complex interval-valued fuzzy sets. Several important mathematical properties are rigorously established. In order to establish the efficacy and implementation of the methods, a real-life application related to pattern recognition involving mineral identification is discussed in detail. Furthermore, based on the proposed correlation and weighted correlation a classification algorithm was developed. The time and space complexities of the proposed algorithm were analyzed. The proposed algorithm was validated through experiments conducted on two real benchmark datasets. The results demonstrate that the proposed approaches were reliable and accurate than several existing approaches by achieving the classification accuracies exceeding 98%.
Complex interval-valued fuzzy sets are powerful tools for representing uncertainty and periodicity that occur in many real-life problems. They not only take both periodicity semantics and uncertainty into account but also describe the information using interval-valued membership grades, which gives experts more freedom to solve complex real-life problems effectively. Complex interval-valued fuzzy sets have been successfully employed many times in medical diagnosis and pattern recognition problems. In this article, two novel methods for computing the correlation coefficient and weighted correlation coefficient of complex interval-valued fuzzy sets are proposed. The methods employed fuzzy statistical parameters such as mean, variance, and covariance of complex interval-valued fuzzy sets. Several important mathematical properties are rigorously established. In order to establish the efficacy and implementation of the methods, a real-life application related to pattern recognition for mineral identification is discussed in detail. Furthermore, based on the proposed correlation and weighted correlation, a classification algorithm is developed. The time and space complexities of the proposed algorithm are analyzed. The proposed algorithm is validated through experiments conducted on two real benchmark datasets. The results demonstrate that the proposed approaches are more reliable and accurate than several existing approaches by achieving classification accuracies exceeding 98%.
Bipolar fuzzy sets provide an effective framework for representing both positive and negative aspects of information. The necessity of a mathematically rigorous and valid distance measure in bipolar fuzzy environments motivates us to introduce a new real-valued function on the set of bipolar fuzzy sets defined over both discrete and continuous universes of discourse. The proposed function is shown to define a valid metric on the set of bipolar fuzzy sets, as it satisfies all the metric axioms. The metric induced by the real-valued function is inspired by the Canberra distance, and it can effectively quantify the dissimilarity between bipolar fuzzy sets in a normalized and interpretable manner. The practical utility of the proposed metric is demonstrated in a pattern recognition problem, where it successfully recognizes an unknown pattern using known bipolar fuzzy patterns. Using the proposed metric, a bipolar fuzzy C-means clustering algorithm is developed for sentiment analysis. The time complexity of the aforementioned algorithm is also analysed. Experiments conducted on the IMDb Movie Review Dataset demonstrate that the proposed algorithm outperforms k-means, fuzzy C-means, and intuitionistic fuzzy C-means algorithms. The proposed bipolar fuzzy C-means algorithm achieves an accuracy of 90.04%, a precision of 90.51%, a recall of 89.01%, an F1-score of 89.75%, a Root mean square error of 0.1191, and a Silhouette score of 0.75. The findings establish that the proposed metric and the associated bipolar fuzzy clustering approach provide a robust and effective framework of handling sentiment data associated with simultaneous positive and negative opinions.
30 March 2026 in Communications in Mathematics and Applications
In pattern recognition, identifying the unknown pattern is a challenging process. Analysing features of the pattern is the initial phase of a pattern recognition process. Features that occasionally show bipolarity have been used for identification. Since Bipolar Fuzzy Sets (BCFSs) cover both the negative and positive features of a pattern in particular, they are utilized to address this bipolarity. BCFSs are one kind of bipolar fuzzy set that produces more accurate results than others. A novel method of pattern recognition is presented here by measuring the correlation coefficient of BCFSs, as BCFSs can handle complex fuzzy information more efficiently. In this study, a method of correlation coefficient of BCFSs is proposed. The pattern recognition method is proposed in a bipolar complex fuzzy environment to resolve uncertain and ambiguous information of an unknown pattern based on the aforementioned approach. A real-life example of the recognition of carbon allotrope is used to validate the efficacy and implementation of the proposed approach.