No / Case-Based Reasoning systems retrieve and reuse solutions for previously solved problems that have been encountered and remembered as cases. In some domains, particularly where the problem solving is a classification task, the retrieved solution can be reused directly. But for design tasks it is common for the retrieved solution to be regarded as an initial solution that should be refined to reflect the differences between the new and retrieved problems. The acquisition of adaptation knowledge to achieve this refinement can be demanding, despite the fact that the knowledge source of stored cases captures a substantial part of the problem-solving expertise. This paper describes an introspective learning approach where the case knowledge itself provides a source from which training data for the adaptation task can be assembled. Different learning algorithms are explored and the effect of the learned adaptations is demonstrated for a demanding component-based pharmaceutical design task, tablet formulation. The evaluation highlights the incremental nature of adaptation as a further reasoning step after nearest-neighbour retrieval. A new property-based classification to adapt symbolic values is proposed, and an ensemble of these property-based adaptation classifiers has been particularly successful for the most difficult of the symbolic adaptation tasks in tablet formulation.

Le Miscanthus x giganteus est une culture pérenne, nouvellement produite en Europe et présentant un intérêt fort pour son usage énergétique. Son implantation présage donc une réorganisation territoriale pérenne. Pour anticiper cette réorganisation, de nombreuses études modélisent les dynamiques spatialement explicites de son insertion. Notre thèse se positionne dans ce courant de recherche. Celle-ci vise à proposer un nouveau cadre de modélisation des processus de décision des agriculteurs, permettant la dissémination horizontale (scaling out) de ces processus issus d’études de cas, vers des territoires élargis. Pour cela, la thèse exploite le modèle du raisonnement à partir de cas. Elle articule (i) une démarche d’acquisition de connaissances sur les processus de décision des agriculteurs relatifs à l’insertion territoriale du miscanthus et (ii) la conception et évaluation d’un prototype ad hoc de raisonnement à partir de cas. La phase d’acquisition des connaissances montre que le processus d’insertion territoriale du miscanthus est complexe : celui-ci est étroitement lié aux contraintes parcellaires du territoire. Ces connaissances nous conduisent alors à discuter du choix des variables biophysiques et humaines intégrées à ce jour dans les modèles spatialement explicites. La phase de conception et d’évaluation du prototype de raisonnement à partir de cas montre que le modèle du raisonnement à partir de cas est particulièrement bien adapté pour modéliser un phénomène contextualisé. Evalués sur nos données d’enquêtes, ces résultats nous conduisent à discuter des modalités d’application du prototype sur d’autres bassins de production de miscanthus / Miscanthus x giganteus is the perennial crop, newly produced in Europe. Even if miscanthus is not so heavily produced nowadays, this crop would be of great interest for energy use. However, the allocation of miscanthus could produce a sustainable reorganization of the landscape. Therefore, many studies aim to model the land use change caused by miscanthus, in order to identify sustainable supply areas: our research belongs to this field. In our research, we propose a new framework for modeling decision-making process of farmers, relying on scaling out. More accurately, we propose to use the case-based reasoning model which solves problems based on an analogical reasoning. Then our research is structured: (i) by a knowledge acquisition step about decision-making process of farmers, based on farm surveys, conducted in the Côte d'Or department (Burgundy region) and (ii) by the design and evaluation of an ad hoc prototype of case-based reasoning. On the one hand, results of knowledge acquisition phase show that miscanthus allocation process is complex, more accurately, that miscanthus allocation process is closely related to land constraints, particularly in terms of logistic and environmental preservation of plots. These results lead us to discuss the selection of biophysical and human variables included to the current spatially explicit models. On the other hand, the design and evaluation phase of our prototype shows that case-based reasoning is particularly well suited to model a contextual phenomenon. These results lead us to discuss the modalities for implementing the prototype in other production areas of miscanthus