Mining Frequent Ordered Patterns without Candidate Generation

Mining frequent patterns is an important data mining task and has been widely studied. However, the traditional frequent pattern mining does not involve the ordered problem, which is widely exists in the real world. A lot of papers have been proposed to solve the ordered problem, including sequential pattern mining, item sequences mining, temporal feature extraction, web log study and ordered patterns mining. Most of these papers used an APRIORI-based algorithm hence did not adopt the wonderful ideas and advanced technologies in traditional frequent patterns mining. This paper introduced a data structure called FOP-tree which is a modified version of FP-tree to solve the ordered patterns mining. The performance study shows that the FOP-tree is efficient and scalable for mining both long and short frequent ordered patterns, and is much faster than the traditional APRIORI-bases algorithms on several situations.