A semi-supervised clustering algorithm for data exploration

This paper is concerned with clustering of data that is partly labelled. It discusses a semi-supervised clustering algorithm based on a modified fuzzy C-Means (FCM) objective function. Semi-supervised clustering finds its application in different situations where data is neither entirely nor accurately labelled. The novelty of this approach is the fact that it takes into consideration the structure of the data and the available knowledge (labels) of patterns. The objective function consists of two components. The first concerns the unsupervised clustering while the second keeps the relationship between classes (available labels) and the clusters generated by the first component. The balance between the two components is tuned by a scaling factor. The algorithm is experimentally evaluated.