Affiliations: [a] Innovative Information Industry Research Center, Harbin Institute of Technology Shenzhen Graduate School, Shenzhen, Guangdong, China | [b] Shenzhen Key Laboratory of Internet Information Collaboration, Harbin Institute of Technology Shenzhen Graduate School, Shenzhen, Guangdong, China | [c] Department of Computer Science and Information Engineering, National University of Kaohsiung, Kaohsiung, Taiwan | [d] Department of Computer Science and Engineering, National Sun Yat-sen University, Kaohsiung, Taiwan | [e] Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan | [f] Department of Industrial and Information Management, National Cheng Kung University, Tainan, Taiwan
Correspondence:
[*]
Corresponding author: Chun-Wei Lin, Innovative Information Industry Research Center/Shenzhen Key Laboratory of Internet Information Collaboration, Harbin Institute of Technology Shenzhen Graduate School, Shenzhen, Guangdong, China. E-mail:jerrylin@ieee.org
Abstract: Mining useful information from large databases has become an important
research area in recent years. Among the classes of knowledge derived,
sequential pattern can be applied in many domains, such as market analysis,
web click streams, and biological data. The fast updated sequential pattern
tree (FUSP-tree) algorithm was proposed to update discovered sequential
patterns in incremental mining. However, it must rescan the original
database for maintaining discovered sequential patterns. This study proposes
the PreFUSP-TREE-INS algorithm based on the pre-large concept for
maintaining discovered sequential patterns without rescanning the original
database until the cumulative number of newly added customer sequences
exceeds a safety bound. The execution time for reconstructing the tree when
old or new customer sequences are added into the original database is
reduced by using pre-large sequences. The pre-large sequences are defined by
lower and upper support thresholds that prevent the movement of sequences
directly from large to small and vice versa. Experiments are conducted to
show the performance of the proposed algorithm for various minimum support
thresholds and ratios of inserted sequences.
Keywords: Data mining, sequential pattern, FUSP-tree, large sequence, incremental mining, dynamic databases
