Author:
Chichang Jou
Affiliation:
Tamkang University, Taiwan
Keyword(s):
Hybrid sequential pattern, Pattern growth, Projected position array, Projected support array, Projected database.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Business Analytics
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
Enterprise Information Systems
;
Health Information Systems
;
Sensor Networks
;
Signal Processing
;
Soft Computing
Abstract:
Sequential pattern mining finds frequently occurring patterns of item sequences from serial orders of items in the transaction database. The set of frequent hybrid sequential patterns obtained by previous researches either is incomplete or does not scale with growing database sizes. We design and implement a Projection-based Hybrid Sequential PAttern Mining algorithm, PHSPAM, to remedy these problems. PHSPAM first builds Supplemented Frequent One Sequence itemset to collect items that may appear in frequent hybrid sequential patterns. The mining procedure is then performed recursively in the pattern growth manner to calculate the support of patterns through projected position arrays, projected support arrays, and projected databases. We compare the results and performances of PHSPAM with those of other hybrid sequential pattern mining algorithms, GFP2 and CHSPAM.