P. Povalej, M. Lenič, T. Welzer-Družovec, M. Družovec, and P. Kokol (Slovenia)
Collaborative systems, multiagent systems, intelligent system construction, multimethod approach, knowledge extraction
The automatic constructive induction from large databases is an important technique used in the field of knowledge mining. Extracted knowledge can be used for categorizing, organizing or predictive purposes. One of the problems encountered is how to make a good induction with good generalization and knowledge representation. Although there is a lot of research going on in this field, it is manily focused on a specific method or on a specific combination of those methods. Another issue of induction process is its multi objective goals. In this paper we introduce a multiagent system that uses multiple methods with multiobjective goals to achieve good generalization and high predictive accuracy. This approach unlike other conventional approaches applies different methods on the same knowledge model, expecting that combined methods will overcome the limitations of each single method and produce better results. A comparison performed on seven different problems showed that our approach in general performs better than single methods.
Important Links:
Go Back
