Influence value defuzzification method

This paper describes a computationally efficient method of defuzzification which we call the influence value (IV) algorithm. The algorithm is meant to be used in conjunction with standard max-min (Mamdani) inference and significantly mitigates the computational cost associated with the aggregation operation followed by the determination of the center of area (COA) of the aggregated output. A straightforward computation of the COA requires a point-by-point maximum operation over the full range of the output variable and then determination of the COA by integration. In contrast, the present method exploits precalculation to achieve an excellent approximation with substantially fewer computations. For practical applications, this permits max-min inference to be carried out with inexpensive processing devices at relatively high speed.