An Approach for the Extraction of Classification Rules from Fuzzy Formal Contexts

This work describes experiments carried out using fuzzy formal concept analysis for the generation of fuzzy classification rules to be used by a genetic process. These rules are simply the intention of the formal concepts extracted from a fuzzy-based formal context. The motivation we have is the need for a method to generate fuzzy classification rules to be used as the search space of the genetic process with reasonable computational cost. The hypothesis we work on is that the extraction of classification rules from a fuzzy formal context can be a suitable alternative to generate the search space of a genetic process due to its low complexity and straightforwardness/simplicity, generating not only good rules, due to the fact that the support of these rules is a subproduct of the extraction of them, but also better interpretable rules, with a varying number of conjunctions in their antecedent part. In order to reduce the complexity of the extraction of the formal concepts, besides the generation of all existing formal concepts representing classification rules, an alternative approach is also proposed and studied: the use of the lattice in order to sequentially extract classification rules. Both ideas are detailed together with the preliminary experiments and results.