Decision Tree Classification Algorithm Based On Rough Set Theory

In data mining study field, decision tree is a regular classification model. Most of decision tree module use univariate test attribute now, but the univariate decision tree module caused some questions of large scale and complex rules that are difficult to understand.In this paper, we introduced simply the decision tree algorithm and the rough set theory, then we present a new multivariate decision tree algorithm based on the rough set theory. By this algorithm, we first present a concept that descript the correlation between the condition attributes, then construct the attribute clusters originally through the concept, after that, we process further the clusters according to the contribution of distinguishing the decision attribute, the clusters are used as new candidate test attributes. In the aspect of selecting test attribute, we use the rules of Variable Precision Weighted Rough Set and merge the sub-branch with a new method. Finally, experiments on the UCI data set show that the decision tree model produced by our algorithm has a good accurate rate and is more simple than ID3 classification models, additionally, the result rules is also easier to be comprehend.