Variable selection and updating in model-based discriminant analysis for high dimensional data with food authenticity applications

摘要

Food authenticity studies are concerned with determining if food samples have been correctly labelled or not. Discriminant analysis methods are an integral part of the methodology for food authentication. Motivated by food authenticity applications, a model-based discriminant analysis method that includes variable selection is presented. The discriminant analysis model isfitted in a semi-supervised manner using both labeled and unlabeled data. The method is shown to give excellent classificationperformance on several high-dimensional multiclass food authenticity datasetswith more variables than observations. The variables selected by theproposed method provide information about which variables are meaningful for classification purposes. A headlong search strategy for variable selection is shown to be efficient in terms ofcomputation and achieves excellent classification performance. Inapplications to several food authenticity datasets, our proposedmethod outperformed default implementations of Random Forests, AdaBoost, transductive SVMs and Bayesian Multinomial Regression by substantialmargins.