Affiliations: [a] Department of Automation, Shanghai Jiao Tong University, Key Laboratory of System Control and Information Processing, Ministry of Education of China, Shanghai, China | [b] Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
Correspondence:
[*]
Corresponding author: Shan Han, Department of Automation, Shanghai Jiao Tong University, Key Laboratory of System Control and Information Processing, Ministry of Education of China, Shanghai 200240, China. Tel.: +86 15000227222; Fax: +86 21 34204305; E-mail:hanshan@alumni.sjtu.edu.cn
Abstract: Incomplete information systems with missing or unknown data may affect the
quality of data-driven decision fusion directly. The impact of missing data
and how much missing data is acceptable for a reliable decision-making
become more important in the era of big data. This paper recommends the
rough set theory for the decision fusion of incomplete information systems
and proposes a new approach to evaluate the impact of missing data.
According to the connection degree tolerance relation, an improved metric
called α-classification quality of approximation is defined to
measure the quality of decision fusion with various identical degrees (IDs).
Then, the link between the volume of missing data and the quality of
decision fusion is established. Furthermore, the relaxed connection degree
tolerance relation is modified to reveal the impact of missing data in the
classification, which makes the influence of changes in the volume of
missing data become assessable. Thus, the assessment method of missing data
is established. The experimental results have shown that the quantitative
evaluation of missing data in an existing information system can be made by
the proposed method and the volume of acceptable missing data according to a
determined quality is possible to be predicted in future applications.
Keywords: Decision fusion, rough set theory, missing data, assessment, method
