Tuning the linear membership functions in spreadsheets to improve the quality of multi-factor fuzzy inference

This paper introduces and illustrates a method of using general-purpose spreadsheets to do fuzzy inference with multi-criteria, when a fuzzy problem has linear membership functions. Three fuzzy sets (labels) represented by eight pairs of parameters are used to partition the domain of a variable. Once these are defined, a spreadsheet can automatically perform the fuzzy inference on any input. A macro is written in MS-Excel for the convenience of implementation. With the help of spreadsheets, two kinds of tune-up are investigated to improve the quality of fuzzy inference. The first is to use membership functions with pair-wise overlap, as opposed to one that uses full overlap (represented here by membership function with three-label overlap). The second is to reshape the membership functions through tuning the medium labels. The results proved that they both help to make the inference more effective and efficient.