Preparing the Aid System for Industrial Network Oriented Diagnostics - New Contributions in Information Systems and Technologies - page 273

The idea is simple and clever: let's build the flat figure on vertex of segments
(vertical position) which represents the sum of identical expert's answers, and build
those segments (horizontal position) in the range [0,1] in position of collected value
of NP(r). As a sample let's consider the situation shown on figure 2, where we have
the flat figure circumscribe based on opinion of 10 experts. One marked “I rather
disagree”, another “I do not have opinion”, another one “I rather agree”, five of
experts marked “I surely agree”, and two others “I absolutely agree”.

Fig. 5. Sample figure - new operator of aggregation of the expert opinion based on the calculation
of the center of gravity of the flat figure

After calculation of the center of gravity of this flat figure we get convincing
degree of correctness of the rules NP agfp (r) (which we assign to X coordinate - x c )
equal to: 0,70353, and the other new parameter called degree of usability of the rule
SP(r) (assign to Y coordinate - y c ) equal to: 1,20908. Using this new parameter SP(r)
we are able to proceed with automation of the choosing the rules for the reasoning
process - simple for computer implementation.
Typically the existing expert systems use so called Certainty Factor - CF , so to be
able to assign values of our new shown operator of aggregation NP agfp (r) to the CF,
which values are in range of [0,1], where 0 means false rule, and 1 means true rule,
we have to change the edge conditions, by closing it both side like that:

⊧ NP ≤ 0 04 ₒ CF = 0

arfp

(1) ⊪ ⊨

NP ⇒ CF 0 041 ≤ NP ≤ 0 95 ₒ CF = NP

arfp arfp arfp

⊪ ⊩

NP ≥ 0 96 ₒ CF = 1

arfp