D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Dealing with Known Unknowns: A Goal-based Approach 15<br />
– EM is the original enterprise model,<br />
– R o is set of observable evolution rules.<br />
– R c is a set of controllable rules applying to the original enterprise model and<br />
other evolved ones.<br />
– Dep ⊆ R o×R o×N is a set of evolution-dependent relations between observable<br />
rules.<br />
R o = [ j r oα = [ j ffff<br />
SM α pα i<br />
−−→ SMi<br />
α<br />
R c = [ n<br />
r cαi = [ nSM α i<br />
Dep = [ n<br />
r oα<br />
o<br />
i<br />
r oβ<br />
where SM α ⊆ EM ∨ ∃SM α′ p α′<br />
i<br />
−−→ SMi α′ .SM α ⊆ SMi<br />
α′<br />
oo<br />
∗<br />
−→ SMij<br />
α<br />
The calculation of the max belief and residual risk for a given configuration<br />
C on a given evolutionary enterprise model eEM is complicated by the need to<br />
consider both multi-part and multi-step evolutions.<br />
However, once we have defined a semantics for the basic set-up of probabilities,<br />
we can now leverage on the mathematics behind the theory or probabilities.Thus<br />
we can use the mechanisms of conditional probabilities for multi-step<br />
evolution and of independent combination of events for multi-part evolution.<br />
The max belief and residual risk of a given configuration C and an evolutionary<br />
enterprise model eEM〈EM, R o, R c, Dep〉 are defined as follows.<br />
(3)<br />
MaxB(C) <br />
max<br />
P r<br />
∀( V α SMα −−→SM pα i<br />
i α). S α SMα i ∈SDA(C)<br />
RRisk(C) 1 −<br />
X<br />
∀( V α SMα pα i<br />
−−→SM α i ). S α SMα i ∈SDA(C) P r<br />
^<br />
SM α pα i<br />
−−→ SMi<br />
α<br />
α<br />
!<br />
^<br />
SM α pα i<br />
−−→ SMi<br />
α<br />
α<br />
! (4)<br />
8 An Incremental Algorithm to Calculate Max Belief and Residual Risk<br />
Developing an algorithm to calculate max belief and residual risk, as in Formula 4,<br />
for an evolutionary goal model is not practical. With the heuristic assumption that<br />
observable rules in different parts are independent we can efficiently eliminate<br />
recursion by transforming the model into a suitable hypegraph and use efficient<br />
hyperpath algorithms.<br />
An evolutionary goal model is converted in to an evolutionary hypergraph as<br />
follows.<br />
Definition 7 (Evolutionary hypergraph) The hypergraph G〈V, E〉 of an evolutionary<br />
goal model 〈EM, R o〉 is constructed as follows:<br />
• For each goal g, create a goal node g.