A Centroid Auto-Fused Hierarchical Fuzzy c-Means Clustering

Like k-means and Gaussian mixture model (GMM), fuzzy c-means (FCM) with soft partition has also become a popular clustering algorithm still is extensively studied. However, these algorithms their variants suffer from some difficulties such as determination of the optimal number clusters which key factor for quality. A common approach overcoming this difficulty to use trial-and-validation strategy, i.e., traversing every integer large like ?n 2 until finding corresponding peak value cluster validity index. But it scarcely possible naturally construct an adaptively agglomerative hierarchical structure using strategy. Even if possible, existing different indices lead clusters. To effectively mitigate problems while motivated by convex clustering, in article we present centroid auto-fused method (CAF-HFCM) whose optimization procedure can automatically agglomerate form hierarchy, more importantly, yielding without resorting any Although recently proposed robust-learning (RL-FCM) obtain best help index, so-involved three hyperparameters need adjust expensively, conversely, our CAF-HFCM involves just one hyperparameter makes adjustment relatively easier operational. Further, additional benefit objective, reduces sensitivity initialization performance. Moreover, able be straightforwardly extended various FCM. Finally, extensive experiments on both synthetic real data sets demonstrate effectiveness efficiency CAF-HFCM.