Uncertain Single Batch Test Algorithm on Cost-sensitive Decision Tree for Uncertain Data
Liu Ming-jian · Gongcheng shuxue xuebao · 2012
Uncertain data wildly exist in various real-world fields.The research community pay more and more attention to investigating data uncertainty.The present cost-sensitive decision tree for uncertain data(CSDTU) can only handle simple test,whose excessive tests result in wasting too much cost.In this paper,we extend the single batch test algorithm of cost-sensitive decision tree for certain data,to CSDTU.We design a new uncertain single batch test algorithm by calculating cost in attribute-selecting of tree with probabilistic cardinality of attribute’s value in uncertain data model.Experimental results on UCI datasets demonstrate that the proposed uncertain single batch can effectively reduce total cost for test and improve the classifier’s performance.It illustrates that uncertain single batch,which is better than simple batch,has good rationality and applicability.