International Journal of Engineering and Innovative Technology (IJEIT)
In this paper, the authors describe a novel approach for inducing fuzzy rules from numerical data. Multidimensional interpolation is employed to construct an interpolation grid, fit to the scattered data. This enables them to automatically obtain a complete fuzzy rule base by associating a fuzzy rule to each point of the grid. Their proposed approach is based on singular spectrum analysis to efficiently reduce noise and improve smoothness of interpolation. Finally, the derived fuzzy rules define a rule-based fuzzy inference system. An application example is provided where it is shown that the estimations are improved.