A More Effective Data Mining Approach That Adroitly Combines Rough Set Theory with Evidence Theory

Longqiang Ni,Zhentang Zhou,Shesheng Gao

DOI: https://doi.org/10.3969/j.issn.1000-2758.2010.06.022

2010-01-01

Abstract:Aim.The introduction of the full paper reviews Refs.1 through 5 and then proposes what we believe to be a more effective data mining approach,which is mathematically explained in sections 1 and 2.The core of section 1 is that we use the rough set theory to reduce the decision table and eliminate redundent information,thus obtaining more concise decision rules.The core of section 2 consists of:(1) we conduct the numerical processing of the decision table thus reduced to cut down the prediction cost;(2) we construct the evidence elements of the combined space with eq.(3),using the coverage factor obtained from the rough set thoery;(3) we apply the evicdece combination rules proposed in subsection 2.3 to synthesizing the conflicting evidences with eq.(7).Section 3 gives a numerical example of fusing the decision rules.The fusion results,presented in Table 4,show preliminarily that our data mining approach,which we believe combines adroitly rough set theory with evidence theory,can effectively refine decision rules and describe them quantitatively.