Chinese Journal of Computers   Full Text
Title	ROLAP-Oriented Temporal Vertical Partitioning Method Based on Rough Sets
Authors	LI Wen-Hai1),2) FENG Yu-Cai2) MA Xiao-Ming3),4) FU Quan2) HU Wen-Bin1)
Address	1)(Department of Computer Engineering, Computer School, Wuhan University, Wuhan 430079) 2)(Institute of Database and Multimedia Technology, College of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074) 3)(Economics and Management School of Wuhan University, Wuhan 430072) 4)(Hubei State Administration of Taxation, Wuhan 430071)
Year	2008
Issue	No.7(1109—1121)
Abstract &   Background	Abstract As the essential element of DSS applications, data warehouse coupled with ROLAP provides an administrant platform with high performance. Especially, it exhibits great efficiency in model construction, data access and operation. Conventional access method of the relational data warehouse is, however, faced with new challenges due to the poor I/O efficiency in ROLAP. In this paper, the characters of DSS are analyzed and the temporal-locality attributes accessed by the operators are proposed in terms of the temporal operations. After this, the corresponding mapping matrix of the temporal access is established by parsing the query template set. With the rough set model, the model PD takes the relational attributes for the clustering objects and regards the queries as the vector attributes. With the proposition of rough clustering algorithm, a series of partitions are obtained through a measurement named effectual benefit criteria under different indiscernibility thresholds. At last, the partitions are evaluated and the storage scheme derived from the most effective partition is designed for the vertically partitioning sub-tables of the relations. Experiments show the proposed scheme is superior to other optimization strategies. Keywords relational operator; temporal clustering; relational storage; indiscernibility degree; rough set model Background This work is a part of the "Research on Automated Storage Design of Relation Database System", which is mainly supported by the National High Technology Research and Development Program (863 Program) under grant No.2004AA4Z3020. The project mainly focuses on the design and development of high-performance relational database management system. The research group including the authors of this paper has proposed several methods concern to some extent areas such as index selection, relational storage and view update based on rough set. With the increasing requirements of mass data storage in DSS applications, high performance storage subsystems are expected to be adaptive with the query templates. In pursuit of this, various strategies and algorithms for efficiently storing and retrieving records have been developed to enhance the performance of the underlying data warehouse system. Rough set has attracted much research interests in the domain of clustering analysis. As for the DSS implement, the authors take relation database as the storage and query bases, and propose an indiscernibility based clustering method to automated partition the fact tables in terms of the query template set.