Encoding symbolic features in simple decision systems over ontological graphs for PSO and neural network based classifiers

Abstract

In the paper, we present the idea of encoding symbolic features appearing in simple decision systems over ontological graphs for building classifiers based on Particle Swarm Optimization (PSO) as well as Neural Networks . Simple decision systems over ontological graphs refer to a general trend in computations proposed by Zadeh and called “computing with words”. In case of such decision systems, we deal with attribute values, describing objects of interest, which are concepts placed in semantic spaces expressed by means of ontological graphs. Ontological graphs deliver us some additional knowledge which can be useful in classification processes. Symbolic data, in our approach in the form of concepts from ontologies, require special treatment to be used in classifiers based on searching for the numerical mapping functions between the known inputs and the corresponding known outputs. • Encoding symbolic features in decision systems over ontological graphs is shown. • Classifiers based on numerical mapping functions can be used for encoded data. • The idea is based on characteristic functions defined according to data semantics. • Encoding based on fuzzy characteristic function becomes a promising method.