Imprecise Gaussian discriminant classification

Gaussian discriminant analysis is a popular classification model, that in the precise case can produce unreliable predictions of high uncertainty (e.g., due to scarce or noisy data). While imprecise probability theory offers nice theoretical framework solve such issues, it has not been yet applied analysis. This work remedies this, by proposing new based on robust Bayesian and near-ignorance priors. The model delivers cautiouspredictions, form set-valued class, limited imperfect available information. We present discuss results experimentation real synthetic datasets, where for this latter we corrupt test instance see how our approach reacts non i.i.d. samples. Experiments show including an component produces reasonably cautious predictions, correspond instances which performs poorly.