A Novel Attribute Reduction Method Using Chi Square Statistics

WEI Li-li,HAN Chong-zhao
DOI: https://doi.org/10.3969/j.issn.1006-9348.2007.05.021
2007-01-01
Abstract:Rough set data analysis has recently become a routine method in categorical data analysis.One of the important problems in rough set theory is attribute reduction.Rough set data analysis doesn't need any other information but data set,simultaneously,the results obtained are usually short of statistical evidence.In this paper,nonparametric methods in statistics are introduced to analyze information systems.The raw information systems is translated into a series of corresponding contingency tables between two attribute subsets,then a new measure of attribute correlation using Chi square statistics is introduced.Based on this measure,a new method of attribute reduction of information systems is presented.The experiments show that the approach proposed is feasible and efficient.
What problem does this paper attempt to address?