Decision Tree Construction based on Rough Set Theory under Characteristic Relation

摘要

Several approaches based on rough set have been proposed for constructing decision tree in complete information systems. In fact, many information systems are incomplete in practical applications. In this paper, a new algorithm, Decision Tree Construction based on Rough Set Theory under Characteristic Relation (DTCRSCR), is proposed for mining classification knowledge from incomplete information systems. Its idea is that the attribute whose weighted mean roughness under the characteristic relation is the smallest will be selected as current splitting node. Experimental results show the decision trees constructed by DTCRSCR tend to have simpler structures and higher classification accuracy.