Performance improvement of the fuzzy rule interpolation method LESFRI

Fuzzy rule interpolation (FRI) methods could be advantageous tools for fuzzy inference owing to their capability to reason in sparse rule bases as well. The fuzzy rule interpolation by the least squares method (LESFRI) like most of the available techniques is designed so that it can be implemented easily only by using the traditional structure based fuzzy inference system (FIS) representation also applied by the FRI Matlab ToolBox. In this paper, we propose a new vector based internal FIS representation (VFIS) and an enhanced version of LESFRI (VLESFRI) designed for the use of VFIS. The performance improvement achievable by the application of VLESFRI is proved by several test results.