Suffix-DDs: Substring Indices Based on Sequence BDDs for Constrained Sequence Mining

In this paper, we study an efficient index structure, called Suffix Decision Diagrams (SuffixDDs), for knowledge discovery in large sequence data. Recently, Loekito, Bailey, and Pei (KAIS, 2009) proposed a new data structure for sequence data, called Sequence Binary Decision Diagram (SeqBDD), which is an extension of Zero-suppressed Binary Decision Diagrams (ZDDs) for sequences. SuffixDD is a compact substring indices based on SeqBDD for efficiently representing the set of all substrings of a given string. Furthermore, SuffixDD provides a rich collection of operations for sets of sequences inherited from ZDDs and SeqBDDs, which are useful for implementing sequence mining algorithms. Then, we present an efficient algorithm for constructing a SuffixDD for a given text, and then, we show the correctness and the complexity. Furthermore, we present a set of extended operations for manipulating SuffixDDs with application to constrained sequence mining. Finally, we give experimental results on the efficiency of SuffixDD.