Enhancing the electre decision support method with semantic data
- Autores: Miriam Martínez Garcia
- Directores de la Tesis: Antonio Moreno Ribas (dir. tes.), Aïda Valls Mateu (dir. tes.)
- Lectura: En la Universitat Rovira i Virgili ( España ) en 2018
- Idioma: español
- Tribunal Calificador de la Tesis: Núria Agell Jané (presid.), Domènec Puig Valls (secret.), Vicenç Torra Reventós (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería Informática y Matemáticas de la Seguridad por la Universidad Rovira i Virgili
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
The field of multi-criteria decision aid (MCDA) studies the way in which people reach a decision when options are defined on a set of criteria, with the aim of developing tools that help the decision maker in this process. In our work, we focus on outranking methods.
This thesis is focused on the methodology ELECTRE (ELimination and Choice Expressing REality) that was, in fact, the first outranking method in MCDA. Outranking methods consider heterogeneous criteria to evaluate the performance of the alternatives and compare them, including different numerical and ordinal scales in the set of criteria. Nowadays it is becoming more common to find decision problems involving non-numerical information, such as multi-valued semantic criteria, which may take as values the concepts of a given domain ontology.
In this PhD Thesis, I propose a new way of handling semantic criteria to avoid the aggregation of the numerical scores before the ranking procedure. This method, called ELECTRE-SEM, follows the same principles than the classic ELECTRE but in this case the concordance and discordance indices are defined in terms of the pairwise comparison of the interest scores.
I also propose to create a semantic user profile by storing preference scores into the ontology. This preferential information may be later exploited to rank and recommend the most suitable alternatives for each user. The numerical interest score attached to the most specific concepts permits to distinguish better the preferences of the user, improving the quality of the decision by the incorporation of an aggregation methodology to infer the user's preferences by considering taxonomic relations between concepts.
The proposed methodology has been applied in two case studies: the assessment of power generation plants and the recommendation of touristic activities in Tarragona.
