Tutorial II genetic fuzzy systems and its application to data mining

Summary form only given. The main objectives of data mining are the accuracy and interpretability of the obtained knowledge from databases. How to design sophisticated algorithms to deal with them is an attractive issue for researches. In recent years, many statistical methods have actively been studied to pursue more and more accurate knowledge. However, the interpretability of such knowledge is often low because of its complexity. On the contrary, symbolic methods have also been studied to obtain interpretable knowledge for further analyses. One of the promising methods is fuzzy data mining which can obtain linguistically interpretable fuzzy if-then rules from databases. To enhance the generalization ability for unseen data, the interpretability for users, and the applicability to real-world problems, evolutionary computation has successfully been incorporated into fuzzy data mining. This approach is often referred to as genetic fuzzy systems. In this tutorial, we introduce the history of genetic fuzzy systems and basic taxonomy. Then we explain two main streams in genetic fuzzy systems: classification and association rule mining. Some current topics such as multiobjective optimization and parallel implementation are also explained.