Model-based classification via mixtures of multivariate t-factor analyzers
A new model-based classification technique is developed, based on the mixtures of multivariate 't'-factor analyzers model and a variation of the EM algorithm is used to estimate model parameters. A convergence criteria based on Aitken's acceleration is used and model selection is carried out using the Bayesian information criteria. This new classification technique is applied to physical and chemical measurements of red wine samples from Italy and to fatty acid measurements on olive oil samples from Italy. These results are discussed and compared to another well-established classification technique. The new method outperformed the traditional technique in virtually all analyses performed.