Clustering data into fuzzy or vague categories is an essential task for data mining and information retrieval applications. In this paper, the authors propose adapting the classical fuzzy C-means (FCM) algorithm to vague sets, an extension of fuzzy sets. In particular, they adapt the optimized function so that both true membership and false membership grades are taken into account, and they design a process based on a variant of particle swarm optimization.

However, this paper shows disturbing elements that limit its contribution. First, there are errors in the preliminary formulas. Second, the comparison is only done on three simple artificial datasets and one “real” dataset. The so-called real dataset is Iris, an overused dataset that no longer relates to what can nowadays be called “real.” Besides, the datasets are described using two or three dimensions, which considerably limits the scope of such a work. Third, if you do not consider the basic FCM algorithm, the observed improvement for the accuracy is about one percent. This does not support the authors’ claim that their approach is “much better” than the other existing approaches.