Módulos de adaptación y retención de un razonador basado en casos
Archivos
Fecha
2018-06-15
Autores
Riverol Quesada, Amanda
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central “Marta Abreu” de Las Villas
Resumen
La empresa de Tecnología de la Información y la Automática, perteneciente a la Unión Eléctrica, desarrolla para el sector eléctrico un Sistema de Información Geográfica denominado SIGOBE. Estudios realizados determinaron la necesidad de un modelo que permitiera el manejo de datos sobre un esquema conceptual del dominio, capaz de dar respuestas a las diferentes peticiones del usuario por medio de consultas automáticas, como soporte a la toma de decisiones. Para la obtención de estas consultas automáticas se implementó, utilizando un enfoque basado en casos, un sistema inteligente de consultas (SICUNE). En el modelo propuesto la base de casos almacena consultas estáticas registradas en el SIGOBE, las cuales se describen por ocho rasgos predictores y tres rasgos objetivos. En la etapa de reutilización, se realiza una adaptación inicial, donde se implementan reglas que contribuyeron al perfeccionamiento de un rasgo objetivo, mejorando notablemente el error de generación de este rasgo, no así con los restantes rasgos, influyendo esto en las soluciones generadas por el sistema, dado que la solución general está conformada por los tres rasgos objetivos. Para dar solución a esta insuficiencia en este trabajo se implementan los módulos de adaptación y retención para el razonador basado en casos que minimizan el error cometido en las soluciones brindadas por el recuperador. Los módulos desarrollados permiten reusar, adaptar usando analogía transformacional y retener las soluciones, teniendo en cuenta los rasgos predictores de tipo ontológico y las definiciones propias del lenguaje SQL. Los resultados evidencian una notable disminución del error en la solución general obtenida. Este sistema es aplicado en todas las instancias de la empresa eléctrica del país.
The company of Information Technology and Automation, belonging to the Electric Union, develops a Geographic Information System called SIGOBE for the electricity sector. Studies conducted determined the need for a model that would allow the management of data on a conceptual scheme of the domain, capable of responding to the different requests of the user through automatic query, as a support to decision making. In order to obtain these automatic queries, an intelligent query system (SICUNE) was implemented using a case-based reasoning. In the proposed model, the case database stores static queries registered in the SIGOBE, which are described by eight predictive features and three objective features. In the reuse stage, an initial adaptation is carried out, where rules are implemented that contributed to the improvement of an objective feature, significantly improving the error of generation of this feature, but not with the other features, influencing this in the solutions generated by the system, since the general solution is made up of the three objective features. To solve this insufficiency in this work, the adaptation and retention modules are implemented for the reasoner based on cases that minimize the error committed in the solutions provided by the recuperator. The developed modules allow to reuse, adapt using transformational analogy and retain the solutions, taking into account the predictive features of ontological type and the definitions of the SQL language. The results show a notable decrease in the error in the general solution obtained. This system is applied in all instances of the country's electricity company.
The company of Information Technology and Automation, belonging to the Electric Union, develops a Geographic Information System called SIGOBE for the electricity sector. Studies conducted determined the need for a model that would allow the management of data on a conceptual scheme of the domain, capable of responding to the different requests of the user through automatic query, as a support to decision making. In order to obtain these automatic queries, an intelligent query system (SICUNE) was implemented using a case-based reasoning. In the proposed model, the case database stores static queries registered in the SIGOBE, which are described by eight predictive features and three objective features. In the reuse stage, an initial adaptation is carried out, where rules are implemented that contributed to the improvement of an objective feature, significantly improving the error of generation of this feature, but not with the other features, influencing this in the solutions generated by the system, since the general solution is made up of the three objective features. To solve this insufficiency in this work, the adaptation and retention modules are implemented for the reasoner based on cases that minimize the error committed in the solutions provided by the recuperator. The developed modules allow to reuse, adapt using transformational analogy and retain the solutions, taking into account the predictive features of ontological type and the definitions of the SQL language. The results show a notable decrease in the error in the general solution obtained. This system is applied in all instances of the country's electricity company.
Descripción
Palabras clave
Razonamiento Basado en Casos, Adaptación, Retención, Inteligencia Artificial