
A Clustering Approach t o Monitor System Working 123
strategies, the construction of an accurate model repre senting the system normal operation
is a nontrivial and important issue.
This chapter presents a clustering approach to monitor sy stem working and detect
changes over time. The main novelty in this approach concerns the change detection strat-
egy that is based on the clus tering extension, i.e., individual memberships reflecting the
current system state. The changes are measured by the disagree ment level between two
partitions obtained from a same temporal data window. The proposed approach is posi-
tioned in a higher abstraction level of