A Novel Approach to Mine Temporal Association Rules

Given a large temporal transaction database, the aim of this paper is to discover the itemsets that are having related support during a particular event over time and to find the association between those items. Most works in mining association rules involve the generation of frequent items which is the core process in generating association rules. It is necessary to scan database in each timeslot to generate frequent items in temporal data mining. This incurs much cost when the number of transactions is large. In this paper, we propose an approach that utilizes the concept of tight lower and upper bounds of supports at different time intervals. It reduces the number of candidates to be scanned in database. Also our method helps in finding association between items that gets related support when a particular event occurs. Our experimental results proves that the association rules generated from the candidate items are more accurate and incurs less cost than the traditional rule mining methods.