Vine copula mixture models and clustering for non-Gaussian data

The majority of finite mixture models suffer from not allowing asymmetric tail dependencies within components and capturing non-elliptical clusters in clustering applications. Since vine copulas are very flexible these types dependencies, we propose a novel copula model for continuous data. We discuss the selection parameter estimation problems further formulate new model-based algorithm. use allows range shapes dependency structures clusters. Our simulation experiments illustrate significant gain accuracy when notably or/and non-Gaussian margins exist. analysis real data sets accompanies proposed method. show that algorithm with outperforms other techniques, especially multivariate