Implementación de algoritmos de asistencia y recomendación para la toma de decisiones y detección de situaciones singulares en sistemas complejos mediante técnicas inteligentes

Abstract

The present research deals with the implementation of tools to provide technical guidance in the decision making, and to develop mechanisms to detect singular events in complex systems. Due to the fact that this Doctoral Thesis is presented following the compendium of publications modality, after an initial contextualization, the content of three research papers, which are published in Journal Citation Reports indexed journals, is included. These research works are linked through a central thread that re ects the progressive evolution of the contributions. Regarding the methods to provide technical guidance, in the rst work it is implemented a system to determine in advance, the anesthetic drug dose in patients undergoing surgery. The developed tool is especially interesting to aid clinical sta about how to proceed during surgery. However, to determine adverse situations, it is emphasized the importance of having a human expert with the ability to identify, from his experience, the normal operation. To face the development of systems to detect singularities without the need a human expert, the second work deals with the use of semisupervised techniques to extract knowledge from the normal performance of a battery. The detection of unexpected events without previous human expertise is carried out with successful results. Finally, the last research work proposes the application of the semisupervised techniques used in the second contribution over the case of study of the rst paper. Then, although the obtained system does not provide an estimation of the patient state, it detects any singular situation in the anesthetic process without the need of a human expert, ensuring the proper surgery development.