Mining Web Frequent Multi-dimensional Sequential Patterns

摘要

Although, numerous methods have been proposed to mine sequential patterns, previous approaches can't effectively find web frequent multi-dimensional sequential patterns from d-dimensional sequence data with multi-dimensional-information, where d>2. The main objective of web frequent multi-dimensional sequential pattern mining is to provide the end user with more useful and interesting patterns. To mine web frequent multidimensional sequential patterns, in present study, we propose a new algorithm ExtSeq-MIDim. It employs extseq (Extended sequential pattern mining method) to mine sequential patterns from d-dimensional sequence data, then forms projected multi-dimensional database for each sequential pattern and uses an algorithm MIDim (Memory Indexing for mining multi-dimensional pattern) to mine multi-dimensional patterns within projected databases. During the multi-dimensional pattern mining process, MIDim takes advantage of the idea of memory indexing without multiple scanning projected databases and handles fewer and shorter multi-dimensional tuples as the discovered patterns get longer. The experimental results show that ExtSeq-MIDim scales up linearly and is efficient to find web multi-dimensional sequential patterns.