Mining Complete Hybrid Sequential Patterns

We discovered that the set of frequent hybrid sequential patterns obtained by previous researches is incomplete, due to the inapplicability of the Apriori principle. We design and implement the CHSPAM algorithm to remedy the problem. CHSPAM first builds the Supplemented Frequent One Sequence itemset (SFOS) to collect items that may appear in a frequent hybrid sequential pattern. It then constructs the projected databases for each item in the SFOS. Its mining procedure is performed recursively in the pattern-growth manner through the projected databases to calculate the support of patterns by backward support counting. We prove the completeness of CHSPAM, and compare the results and performances of CHSPAM with those of GFP2, the most efficient hybrid sequential pattern mining algorithm as far as we know.

• Type: Conference
• Year: 2006
• Where: DMIN
• Authors: Chichang Jou
• Keywords: DMIN 2006, DMIN 2007, Frequent Hybrid, Frequent One Sequence, Hybrid Sequential Pattern