Computational Intelligence

404 Chapter Ten—Performance Metrics

In our test case, even though one of the runs using configuration A had the hig hest

fitness of the nine runs, we cannot say that configuration A is significantly better

than our baseline configuration B to a significance level of 0.05.

Now, let’s test another configuration of the EC tool. Let’s call this configuration

C. As before, we make five runs using configuration C and compare the fitness values

with the four baseline cases using configuration B. Ignoring the specific fitness values

for purposes of this illustration (since the rank is all that matters), we obtain:

C C C C B B C B B

Using the simple and direct method of