Finding Interesting Sequential Patterns in Sequence Data Streams via a Time-Interval Weighting Approach

Joong Hyuk CHANG  Nam Hun PARK  

IEICE TRANSACTIONS on Information and Systems   Vol.E96-D   No.8   pp.1734-1744
Publication Date: 2013/08/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E96.D.1734
Print ISSN: 0916-8532
Type of Manuscript: PAPER
Category: Artificial Intelligence, Data Mining
time-interval weight,  weighted sequential pattern,  time-interval sequential pattern,  time-interval sequence data stream,  data stream,  

Full Text: PDF(1.3MB)>>
Buy this Article

The mining problem over data streams has recently been attracting considerable attention thanks to the usefulness of data mining in various application fields of information science, and sequence data streams are so common in daily life. Therefore, a study on mining sequential patterns over sequence data streams can give valuable results for wide use in various application fields. This paper proposes a new framework for mining novel interesting sequential patterns over a sequence data stream and a mining method based on the framework. Assuming that a sequence with small time-intervals between its data elements is more valuable than others with large time-intervals, the novel interesting sequential pattern is defined and found by analyzing the time-intervals of data elements in a sequence as well as their orders. The proposed framework is capable of obtaining more interesting sequential patterns over sequence data streams whose data elements are highly correlated in terms of generation time.