A reconciliation-driven approach of case-based prediction: state of the art, method overview and application in food science
Résumé
This chapter proposes an approach to generate predictions for decision support issues. It relies on case-based and reconciliation methods, using an ontology. The objective of the chapter is to provide an overview of the state of the art, but also to describe the proposed method and to illustrate it on a concrete application. In this approach, a reconciliation stage identifies groups of rules expressing a common experimental tendency. A prediction stage generates new rules, using both experimental tendencies obtained in the previous stage and new experimental descriptions. The method has been tested within a case study concerning food quality management. It has been compared to a classic predictive approach, leading to promising results in terms of accuracy, completeness and error rate.