An Efficient Algorithm of Constructing Sequence BDDs for Large-Scale Sets of Sequences and Its Evaluation

Abstract

Representing and processing large-scale sets of sequences is an important technique for a wide range of applications such as indexing sequences and data mining. Sequence BDD, SeqBDD, proposed by Loekito et al., is an efficient data structure for representing string sets. We can apply rich operations to SeqBDDs while keeping the set of sequences compressed. In this study, we propose a bottom-up algorithm for construction of SeqBDDs from large text-data which contain a large number of sequences. Compared to the existing algorithm that sequentially add sequences by union operations, the proposed algorithm works faster and more efficiently in many input data sets, especially data sets where many sequences with common prefixes appear. In this paper, we describe the proposed algorithm and evaluate its effectiveness by experiments using various data sets