Attribute reduction on measuring data in neighborhood rough set with common-test-cost and error ranges

作     者：刘忠慧 

作者机构：西南石油大学计算机科学学院 

出 版 物：《科技展望》 (Science and Technology)

年 卷 期：2017年第27卷第18期

页      面：257-261+266页

学科分类：08[工学] 081104[工学-模式识别与智能系统] 0811[工学-控制科学与工程] 

基　　金：part supported by Scientific Research Starting Project (No. 2014QHZ025) the 2nd Young Scholar Training Program of SWPU 

主　　题：Attribute reduction Neighborhood rough set Error ranges Common-test-costs Heuristic algorithm 

摘      要：Many previous studies on measuring data attempted to seek an optimal reduct to achieve a low total test cost,which is based on an assumption that all attributes are independent. In many real-world measurements,however,this assumption is not reasonable due to the affection of some attributes being related. To address this issue,we firstly define decision systems with test-costs,common-test-costs and error ranges for data reduction. The concepts of reduction which include lower and upper approximations,positive regions,and relative reducts are based on neighborhood rough sets. Then we design a heuristic algorithms to attack measuring data sets reduction. In the process of reduction,common-test-costs are used to improve prune strategy and develop information function for algorithm ’s efficiency. The experiments were conducted on 7 UCI measuring database,which validated the effectiveness of proposed algorithms.