An efficient algorithm to discover calendar-based temporal association rules

This work is to discover all calendar-based temporal association rules that may occur over any time interval in a temporal database. A user-given calendar schema, e.g., year, month, and day, is firstly adopted to specify the interesting time intervals as calendar patterns. Then, in every time interval, the frequent 2-itemsets are discovered along with their 1-star calendar patterns. After that, information of the rest k-star calendar patterns of the frequent 2-itemsets are level wisely aggregated from their 1-star calendar patterns. A minimal set of candidate calendar patterns are generated and counted in the first scan of database. To avoid multiple scans over the database, all candidate itemsets are generated from frequent 2-itemsets and the a priori downward property is utilized to reduce the number of candidate calendar patterns. Finally, all frequent itemsets with their frequent calendar patterns are discovered in one shot. Calendar-based temporal association rules are then obtained. Experimental results have shown that our method is more efficient than others.