Discovering Temporal Patterns for Interval-based Events

In many daily transactions, the time when an event takes place is known and stored in databases. Examples range from sales records, stock exchange, patient records, to scientific databases in geophysics and astronomy. Such databases incorporate the concept of time which describes when an event starts and ends as historical records [9]. The temporal nature of data provides us with a better understanding of the trend or pattern over time. In market-basket data, we can have a pattern like “75% of customers buy peanuts when butter starts to be in big sales and before bread is sold out”. We observe that there may be some correlations among peanuts, butter and bread so that we can have better planning for marketing strategy. Knowledge discovery in temporal databases thus catches the attention of researchers [8, 4].