• data mining
• fuzzy logic
• genetic algorithms
• multi-objective genetic algorithms
• fuzzy rule based classifiers

Nowadays, organizations deal with rapidly increasing amount of data that is stored in their databases. It has therefore become of crucial importance for them to identify the necessary patterns in these large databases to turn row data into valuable and actionable information. By exploring these important datasets, the organizations gain competitive advantage against other competitors, based on the assumption that the added value of Knowledge Management Systems strength is first and foremost to facilitate the decision making process. Especially if we consider the importance of knowledge in the 21st century, data mining can be seen as a very effective tool to explore the essential data that foster competitive gain in a changing environment.
The overall aim of this study is to design the rule base component of a fuzzy rule-based system (FRBS) through the use of genetic algorithms. The main objective is to generate accurate and interpretable models of the data trying to overcome the existing tradeoff between accuracy and interpretability. We propose two different approaches: an accuracy-driven single-objective genetic algorithm, and a three-objective genetic algorithm that produce a Pareto front approximation, composed of classifiers with different tradeoffs between accuracy and complexity. The proposed approaches have been compared with two other systems, namely a rule selection single-objective algorithm, and a three-objective algorithm. The latter has been developed by the University of Pisa and is able to generate the rule base, while simultaneously learning the definition points of the membership functions, by taking into account both the accuracy and the interpretability of the final model.