Gesconda

The main goal of the project is to design and develop a prototype of a multi-agent tool for intelligent data analysis and implicit knowledge management of data bases, with special focus on environmental data bases. It is remarkable the high quantity of information and knowledge patterns implicit in large data bases coming from the monitoring of any system or dynamical environmental process, for instance, historical data collected about meteorological phenomena in a certain area, about the performance of a wastewater treatment plant, about characterizing environmental emergencies (toxic substances wasting, inflammable gas expansion), about geomorphologycal description of seismic activity, etc.

All this information and knowledge is very important for prediction tasks, control, supervision and minimization of environmental impact either in Nature and Human beings themselves. The tool that is proposed here will be composed by several statistical data analysis methods (oneway and twoway descriptive statistics, missing data analysis, clustering, relations between variables) as well as several machine learning techniques, coming from Artificial Intelligence (clustering, case based reasoning techniques, reinforcement learning, dynamical analysis).

As main differences from the existing commercial systems, the more relevant aspects of this proposal are: the interaction of the developed methods, the development of mixed techniques that can cooperate among them to extract the knowledge contained in data, the existence of dynamical data analysis, and the existence of a recommender agent which will suggest the best method to be used depending on the target domain and on the goals especified by users.
 
This project is funded by the Spanish Research Council (CICYT), code TIC2000-1011.