Towards a cost-effective and fast traceability assessment: A principal component exploratory analysis

Abstract

We are presenting an exploratory analysis performed in order to assess the feasibility of building a multivariate tool designed to perform a cost-effective and fast traceability assessment. The evaluation has been performed by using Principal Component Analysis (PCA), as this non-supervised artificial intelligence technique reveals the structure of the original data and allows the evaluation of the clustering quality. It also allows an objective variable selection, as it indicates the most important variables which are contributing to the clustering of the data and which variables are redundant and thus may be discarded. The system has been tested for green peas (Pisum sativum), which is one of the most popular vegetable on the European horticultural market. The results show that the proposed PCA system can also be used as a stand-alone tool for traceability assessments, as it allows the assignment of the modeled country of origin by performing a binary (asymmetric) classification. The system is very user-friendly, even for non-specialists such as law enforcement officers, as its graphical interface is easy to understand.