مشخصات پژوهش

Intuitionistic fuzzy c-means clustering (IFCM) is a helpful approach for dealing with uncertain data. However, it still suffers from noise and center initialization, and difficulty discerning the relative impacts of samples, fea￾tures, or clusters during clustering. On the other hand, semi-supervised learning improves clustering by using class-label information that is available only for some data samples. This integration creates semi-supervised clustering, a concept not extensively explored in intuitionistic fuzzy clustering. In light of the challenges mentioned for IFCM, this paper proposes a novel semi-supervised IFCM-based approach. A new semi-supervised intuitionistic fuzzy objective function is presented, which is composed of (1) sample weighting to reduce the effect of outliers, (2) a feature weighting to distinguish between the importance of different features, (3) cluster weighting to reduce the initialization sensitivity, (4) a neighborhood information term reflecting the extent of label homogeneity in the neighboring samples. Considering the above factors in designing the objective function leads to the formation of an optimal clustering structure, insensitivity to noise/outliers, and insensitivity to the initialization. Extensive experiments on various UCI benchmark datasets demonstrate that the proposed method outperforms competing methods by an average accuracy improvement of 11.22 % across all datasets. Further￾more, statistical tests confirm the significance of these improvements. The source code and detailed instructions for reproducing the experiments are available on GitHub at https://github.com/Amin-Golzari-Oskouei/NA-SS IFCM-MLW.