Paper Info
Title
Attribute Importance Measurement Based On The Data Effect And Its Performance Analysis In Computation Practice
Abstract
Attribute importance measurement is the core of multi-attribute decision making. This study aims at resolving the problem of attribute importance computation. Based on the data system and the inclusion degree of sets, we propose a data deletion method based on knowledge reliability. By taking the hidden knowledge in the data system as the carrier, we further discuss the changing rules of knowledge factors in the subsystems and propose an attribute importance measurement method based on the data effect (abbreviated DE-AIM). Furthermore, we analyze the values and structure features of DE-AIM through a theoretical proof and an example calculation. Finally, we compare DE-AIM with other attribute importance measurements by combining with concrete cases and several data sets from UCI. All the results indicate that DE-AIM not only has good structural features and interpretability but also can reflect different decision preference by changing parameters. The method has wide applications in many fields such as resource management, artificial intelligence, complex system optimization, expert systems, and concurrency computation.
Year
DOI
Venue
2021
10.1002/cpe.5038
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
Keywords
DocType
Volume
attribute class, attribute importance measurement, core data, decision information system
Journal
33
Issue
ISSN
Citations 
8
1532-0626
0
PageRank 
References 
Authors
0.34
20
3
Name
Order
Citations
PageRank
Fachao Li100.34
Chenxia Jin210113.20
Xiao Zhang310.69