Table 
shows the results of the experiment. The newsgroup field determines the search
domain of the profile and indicates the kinds of articles it is likely to retrieve.
Hence, mutation only operates on the newsgroup field thereby directing the exploratory
process. The results are shown with different initial profiles.
The initial profile for each test run is shown in the left column. The initial profile is successively mutated over many generations. This process is repeated a large number of times for the same profile to average out the random effects of the mutation operator. So for every initial profile, a probability distribution for the final profile is derived. The distribution gives the probability of a newsgroup being in the final profile.
For each test run the initial profile, the final profile probability distribution
and the number of generations is indicated. Continual mutations produce some
interesting results. The system appears to be exploring in the neighborhood
of the initial profile and coming up with interesting newsgroups that are quite
similar to the ones in the initial profile. The implication is that if the user
likes a newsgroup in the initial profile, say, clari.biz.finance
,
then it is quite likely (with probability ,
see Table
)
that he is also interested in the newsgroup clari.biz.mergers.