Analyzing and extending hierarchical systems of linguistic rules

We propose an approach to design linguistic models which are accurate to a high degree. To do so, we use an extension of the structure of the knowledge base of fuzzy rule base systems, i.e., a more flexible hierarchical knowledge base. This flexibility will allow us to have linguistic rules defined over linguistic partitions with different granularity levels, and thus to improve the modeling of those problem subspaces where the former models have bad performance. The methodology proposed in this paper has been thought as an extension of a previous two-level methodology, two hierarchical levels. First, we extend it by developing an iterative hierarchical systems of linguistic rules learning methodology with the purpose of performing an accurate refinement of linguistic models in each step of an iterative process, more than two levels. Later, we extend the hierarchical rule base structure making it more flexible, reinforcing the action of a rule in the subspace where it is defined. This will be performed by allowing the use of weighted and double-consequent reinforced rules