Develop Multi-hierarchy Classification Model: Rough Set Based Feature Decomposition Method

• Published in: Lecture notes in computer science pp. 164 - 171
• Main Authors: Wang, Qingdong, Dai, Huaping, Sun, Youxian
• Format: Book Chapter Conference Proceeding
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg 2005; Springer
• Series: Lecture Notes in Computer Science
• Subjects: Applied sciences; classification; Computer science; control theory; systems; concept hierarchy; Data processing. List processing. Character string processing; Exact sciences and technology; feature decomposition; Memory organisation. Data processing; Rough set; rule induction; Software; Data analysis; Model-based reasoning; Transparency; High precision; Knowledge base; Information extraction; Pattern recognition; Data mining; Modeling; Rough set theory; Measure theory; Classification; Database; Artificial intelligence; Inference rule
• ISBN: 3540287574; 9783540287575
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/11551188_18

Model development on high dimension database is very difficult. This paper presents a new rough set based machine learning method, named feature decomposition method, to discover concept hierarchies and develop a multi-hierarchy model of database. For the databases which we are familiar with, the feature group can be selected by experience of expert. When dealing with the databases without any background knowledge, a new criterion based on rough set is presented to select the features to form a feature group. According to some measures of rough set theory, the objects defined on the proposed feature group are labeled by a new intermediate concept. The concept hierarchies of the database have specific meaning, which increased the transparency of data mining process and enhance the comprehensibility of the model. Each feature group and the corresponding intermediate concept compose the structure of the database. Finally rule induction can be processed on the intermediate concepts. The algorithm presented is verified by datasets from UCI. The results show that the multi-hierarchy model established by feature decomposition method can get high classification accuracy and have better comprehensibility.