Extracting Connection Types in Process Models Discovered by Using From-to Chart Based Approach - Developing Concepts in Applied Intelligence - page 62

After FROMTOCHART table is created and parsed transitions in activity streams are
populated to this table, scores are evaluated with respect to support and confidence
thresholds ( supportFTC=0.40 and confidenceFTC=0.40 ). Table 1 shows the initial
and final states of FROMTOCHART table.
After scores are evaluated, GA engine reproduces activity sequence starting with
firCon activity and then calculates fitness score for these activity sequences
according to proposed population size ( populationSize=80 ) and maximum number
of iterations ( maxIterationNum=200 ) settings.
As shown in Figure 1, initial population is quite diverse early on the process, so
crossover frequently takes large steps early in the search space and fitness scores of
the individuals tend to increase. Afterwards smaller steps are taken when most
individuals are quite similar, i.e. the gap between average and maximum fitness score
boundaries has shrunk. Consequently the fittest (and optimum) activity sequence is
obtained as { firCon , makTic , arrSur , infCliSur , sur , intRep , repRdy , rdyInfCli ,
senTic , ticRdy } with a 0.779 fitness score.
Fig. 1. Runtime Analysis for Rearrangement Operation.
After rearrangement of FROMTOCHART table, relations are extracted according to
positive scores remarked at FROMTOCHART table. Consequently, dependency/frequency
graph, which visually converts the fittest activity sequence ( genotype ) into the process
model ( phenotype ), is constructed as shown in Figure 2.
Fig. 2. Extracted Relations in Dependency/Frequency Graph.