ABSTRACT

Supervised learning is the process of disposition of a set of consanguine data items which have known labels. The apportion of an unlabeled dataset into a conglomeration of analogous items(clusters) by the optimization of an objective function to attenuate the inter-class similarity and augment the intra-class similarity is called unsupervised learning. But when multi-modal data is used, there ensues a predicament with algorithms of either type. Hence a new breed of clustering known as Semi-Supervised clustering has been popularized. This algorithm partitions an unlabelled data set into a congregation of data items by taking only the limited available information from the user.
When contemporary clustering algorithms are applied on a single dataset, different result sets are obtained. Hence an algorithm is needed to reveal the underlying structure of the dataset. In this paper an algorithm for semi-supervised learning is endowed, quartered on the principle of collaboration of clusters. This analytical study can be justified by carrying out various experiments.