Fuzzy magnetic optimization clustering algorithm with its application to health care

Clustering is an important tool for data mining and knowledge discovery that helps in revealing hidden structures and “clusters” found in large data sets. Fuzzy C-means (FCM) is considered to be popular data clustering method due to its capability of clustering the datasets that are uncertain, vague and/or are otherwise difficult to cluster. Although, noted both for its simplicity of implementation and its output validity, performance of FCM usually gets affected in case of poor initialization resulting in the algorithm getting trapped into a local optimum. To overcome this shortcoming, the present study proposes a novel clustering algorithm called fuzzy magnetic optimization clustering (Fuzzy-MOC) which embeds the concept of fuzzy clustering into magnetic optimization algorithm. In Fuzzy-MOC, the data points apply force directly to the magnetic particles due to which the particles change their positions in the feature space. Magnetic particles are attracted by their neighbours assumed to be in a lattice like structure. The proposed algorithm is evaluated on a set of 16 benchmark datasets taken from the UCI Machine Learning Repository including high dimensional gene expression dataset. Experimental results demonstrate that Fuzzy-MOC outperforms the other state-of-the-art algorithms in terms of different performance metrics like F1, accuracy, purity and RI measure. © Springer-Verlag GmbH Germany, part of Springer Nature 2018.