From data to decision support: A multicriteria model for interpretable dropout risk classification

In this paper, a model for multicriteria ordinal classification is proposed and used for student dropout risk assessment in a structured, understandable, and uncertain manner. The problem is modeled as a multicriteria sorting problem, where a student is evaluated with respect to a set of criteria and then classified according to the risk level. Interval-based performance representations are used to account for the presence of uncertainties and incomplete information, as well as possible interactions between criteria, without using oversimplified additive assumptions. Preference parameters are not fixed a priori but are derived through a disaggregation process using observed data. The model is applied to a sample of 45 students, each representing a unit of analysis. Data collection is conducted using a structured tool, and the results are aggregated into a matrix for evaluation. The classification results provide a clear risk profile for each student, both globally and on a dimensional level. It is not a matter of seeking any generalization but rather illustrating the usefulness of the model as a support for decisions in academic monitoring.