Record reduction based on attribute oriented generalization

Author

Wang, Li-zhen ; Chen, Hong-mei

Author_Institution

Sch. of Inf., Yunnan Univ., Kunming, China

Volume

3

fYear

2005

fDate

18-21 Aug. 2005

Firstpage

1693

Abstract

Record reduction is very important in the research and application of KDD. The aim of record reduction is to keep less record count and more information amount. Ratio of record reduction (RRR) and information amount based on semantic proximity (IABSP) are presented as measures. Record reduction is analyzed from two aspects of rules and measures in order to ensure the correction and effectiveness of results. In this paper, record reduction is materialized as record reduction based on attribute oriented generalization (RRBAOG). A new AOG method based on partition, prune and optimization strategies is presented in order to improve the execution efficiency of RRBAOG. Two algorithms of RRBAOG, from bottom to top (FBTT) and from top to bottom (FTTB) are also given. The efficiency of algorithms is analyzed by experiments.

Keywords

data mining; generalisation (artificial intelligence); IABSP; RRBAOG; attribute-oriented generalization; databases; from bottom to top algorithm; from top to bottom algorithm; knowledge discovery; optimization; record reduction; semantic proximity; Algorithm design and analysis; Application software; Computer science; Data mining; Data preprocessing; Databases; Machine learning; Machine learning algorithms; Optimization methods; Partitioning algorithms; Record reduction; attribute-oriented generalization; information amount based on semantic proximity; record reduction based on attribute-oriented generalization;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning and Cybernetics, 2005. Proceedings of 2005 International Conference on

Conference_Location

Guangzhou, China

Print_ISBN

0-7803-9091-1

Type

conf

DOI

10.1109/ICMLC.2005.1527217

Filename

1527217