Extending a Hybrid CBR-ANN Model by Modeling Predictive Attributes Using Fuzzy Sets

This paper presents an extension of an existing hybrid model for the development of knowledge-based systems, combining case-based reasoning (CBR) and artificial neural networks (ANN). The extension consists of the modeling of predictive attributes in terms of fuzzy sets. As such, representative values for numeric attributes are fuzzy sets, facilitating the use of natural language, thus accounting for words with ambiguous meanings. The topology and learning of the associative ANN are based on these representative values. The ANN is used for suggesting the value of the target attribute for a given query. Afterwards, the case-based module justifies the solution provided by the ANN using a similarity function, which includes the weights of the ANN and the membership degrees in the fuzzy sets considered. Experimental results show that the proposed model preserves the advantages of the hybridization used in the original model, while guaranteeing robustness and interpretability.