Fuzzy C-Means clustering based on dual expression between cluster prototypes and reconstructed data

The Fuzzy C-Means (FCM) algorithm is one of the most commonly used clustering methods. In this study, the reconstructed data supervised by the original data is introduced into the FCM clustering, and a dual expression between cluster prototypes and reconstructed data is mined by extending the FCM clustering model using cluster prototypes, memberships and reconstructed data as variables. The convergence and the time complexity of the proposed algorithm are also discussed. Experiments using synthetic data sets and real-world data sets are focused on the influence of the extent to which the reconstructed data are supervised by the original data on the clustering performance. A way of parameter selection is provided which is helpful for enhancing the usefulness of the proposed algorithm. An application case study for monitoring data of shield construction is also presented. It reveals the effectiveness of the proposed algorithm from the viewpoints of the interpretability of clustering results and the representativeness of cluster prototypes.