Download RAW file: https://library.csbe-scgab.ca/docs/journal/40/40_2_127_raw.pdf
The objective of this research was to develop and test a procedure for the determination of fuzzy membership functions (MF) based on data acquired from experts. The area ofstudy was the fuzzification ofmilk yield on a given test day. The procedure consisted ofshowing 324 test day records (representing 36 herds and 1103 cows) randomly to two experts and asking them to qualify milk yield using one or two consecutive fuzzy sets out of five (very low, low, medium, high, and very high). They were also asked to furnish a degree of membership to each set in the case of two being chosen. The qualification of milk yield was based on its comparison with standard values determined as a function ofparity, days in lactation, and herd average milk yield. The data acquired from experts were first analyzed graphically to I) verify the dependence of these data on parity, days in lactation, and herd average milk yield, and 2) determine the position ofcritical points for the fuzzy sets. The MF were then established by fitting a non-linear model to the data within each region bounded by the critical points. The fuzzy sets appeared to be little influenced by parity, days in lactation, and herd average milk yield. However, important discrepancies were found between the two experts, especially for extreme fuzzy sets. The membership functions also differed from those assessed by the experts prior to this project. Although the method used to determine optimal sets ofmodel parameters for MF may be further improved, the overall procedure appears to be appropriate for rapidly finding the MF which reflect well the thinking ofexperts.
R. Lacroix, R. Tietnessen and K.M. Wade 1998. DETERMINATION OF FUZZY MEMBERSHIP FUNCTIO 'S FOR MILK YIELD THROUGH STRUCTURED I 'TERVIEWING OF EXPERTS. Canadian Agricultural Engineering 40(2):127-137.
Canadian Society for Bioengineering