| 摘要 |
<p>PROBLEM TO BE SOLVED: To detect a causation from cooccurrence between different transactions. SOLUTION: When a causation is detected from database that stores history of events, an event list which sorts events in occurrence time order is generated. The appearance count of an event set pair (X and Y) that has such a cooccurrnece relation in which a 2nd event set Y(=(ey1 , ey2 ,..., eyn ,...), eyn shows an event, and n is a natural number) happens from the occurrence time of a 1st event set X(=(ex1 , ex2 ,..., exn ,...), exn shows an event, and n is a natural number) until the preliminarily defined time (window size w) passes and the appearance counts of the set X of the 1st events and the set Y of the 2nd events are calculated. The significance of causation in which the 2nd event set Y results from the 1st event set X in the window size w is decided by using the appearance count of the event set pair (X and Y) and the appearance count of each event set X and Y.</p> |
