Proposed Bounded Variation Based Probabilistic Intuitionistic Fuzzy C-Means Clustering Algorithm

Euclidean distance measures are mostly utilized in proposing probabilistic variants of local optima-based intuitionistic fuzzy c-mean clustering algorithms. The partitional clustering algorithm involving global optima is missing from the literature. The probabilistic intuitionistic fuzzy c-mean algorithm (PIFCM) is a well-known clustering algorithm converging to a local minima and hence variations in the optimal outcomes are observed. So, researchers are using several types of distance measures to improvise the c-mean algorithm. In this paper, we exploit a step-based distance measure involving bounded variation (BV) and probabilistic weights to propose the bounded variation-based probabilistic intuitionistic fuzzy c-mean algorithm (BV-PIFCM). This approach is particularly effective in datasets where clusters are non-globular or have irregular shapes. Step-Based distance measures nearly capture the realistic scenarios as they move along irregular paths, and hence cluster detection becomes easy even in varying densities or sparse regions. Experimental results on the car data set and multiple benchmark UCI datasets demonstrate that our proposed algorithm outperforms PIFCM and IFCM algorithms. Here, the benchmark indexes used are accuracy, partition coefficients, and cluster entropy as performance metrics.