A system learning user preferences for multiobjective optimization of facility layouts

Áreas de investigación:

• Methodology - Machine learning

Año:

2012

Tipo de publicación:

Artículo en conferencia

Autores:

• Pérez-Ortiz, María
• Arauzo-Azofra, Antonio
• Hervás-Martínez, César
• García-Hernández, Laura
• Salas-Morera, Lorenzo

Título del libro:

Proceedings on the Int. Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO'12)

BibTex:

@conference{multiobjPerez2012, author = "Mar{\'i}a P{\'e}rez-Ortiz and Antonio Arauzo-Azofra and C{\'e}sar Herv{\'a}s-Mart{\'i}nez and Laura Garc{\'i}a-Hern{\'a}ndez and Lorenzo Salas-Morera", abstract = "A multiobjective optimization system based on both subjective and objective information for assisting facility layout design is proposed on this contribution. A dataset is constructed based on the expert evaluation of some facility layouts generated by an interactive genetic algorithm. This dataset is used for training a classification algorithm which produces a model of user subjective preferences over the layout designs. The evaluation model obtained is integrated into a multi-objective optimization algorithm as an objective together with reducing material flow cost. In this way, the algorithm exploits the search space in order to obtain a satisfactory set of plant layouts. The proposal is applied on a design problem case where the classification algorithm demonstrated that it could fairly learn the user preferences, as the model obtained worked well guiding the search and finding good solutions, which are better in term of user evaluation with almost the same material flow cost. ", booktitle = "Proceedings on the Int. Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO{\&}#039;12)", title = "{A} system learning user preferences for multiobjective optimization of facility layouts", year = "2012", }

Abstract:

A multiobjective optimization system based on both subjective and objective information for assisting facility layout design is proposed on this contribution. A dataset is constructed based on the expert evaluation of some facility layouts generated by an interactive genetic algorithm. This dataset is used for training a classification algorithm which produces a model of user subjective preferences over the layout designs. The evaluation model obtained is integrated into a multi-objective optimization algorithm as an objective together with reducing material flow cost. In this way, the algorithm exploits the search space in order to obtain a satisfactory set of plant layouts. The proposal is applied on a design problem case where the classification algorithm demonstrated that it could fairly learn the user preferences, as the model obtained worked well guiding the search and finding good solutions, which are better in term of user evaluation with almost the same material flow cost.