Online Variational Learning of Shifted Scaled Dirichlet Mixture

Clustering is a typical unsupervised learning method applied to discover and analyze group structure in a set of observations. Among all of the clustering methods, mixture models have become one of the principal data mining approaches to explore various types of data and find the hidden structures. In real-time and large-scale applications, data instances are generated sequentially. Due to the nature of such data, online algorithms have grabbed lots of attention. In this paper, we propose a novel online clustering approach based on finite shifted scaled Dirichlet mixture models. We use a variational approach with an online setting for learning our model. This proposed algorithm is validated through two real-world applications related to smart buildings.