A data mining approach to cognitive mapping is presented based on bipolar logic, bipolar relations, and bipolar clustering. It is shown that a correlation network derived from a database can be converted to a bipolar cognitive map (or bipolar relation). A transitive, symmetric, and reflexive bipolar relation (equilibrium relation) can be used to identify focal links in decision analysis. It can also be used to cluster a set of events or itemsets into three different clusters: coalition sets, conflict sets, and harmony sets. The coalition sets are positively correlated events or itemsets; each conflict set is a negatively correlated set of two coalition subsets; and a harmony set consists of events that are both negatively and positively correlated. A cognitive map and the clusters can then be used for online decision analysis. This approach combines knowledge discovery with the views of decision makers and provides an effective means for online analytical processing (OLAP) and online analytical mining (OLAM).
|