Aplicación de métodos de aprendizaje automático en un sistema basado en ontología

The ontology-based system for the management of environmental indicators in corporations (SIGCIA) allows the detection of an indicator alteration, if it exceeds a limit value. In this case, this system recommends the possible environmental impacts, the causes of the indicator alteration and the mitigation actions. In order to make these recommendations, the limit value for each indicator must be pre-defined in the software by the environmental management specialist. This means that the determination of limit values is done subjectively, based on the knowledge of the historical behavior of the indicator in a specific organization; so it is necessary to have an automatic forecast method. This research transits through all the phases of the process of Knowledge Discovery in Data (KDD). A selection of attributes in the dataset was made applying several selectors and a group of regression models were applied. Artificial Neural Networks with Multi-Layer Perceptron topology showed best performance. It allows the prediction of the limit value of the energy consumption indicator, dataset selected as study case. The prediction of limit values and the potential offered by the ontology-based recommendation system make it a powerful tool to support decision-making in the process of environmental management, with broad generalization possibilities in Cuban business sector.