For the moment, the attribute reduction algorithm of relative knowledge
granularity is very important research areas. It provides a new viewpoint to
simplify feature set. Based on the decision information is unchanged, fast
and accurate deletion of redundant attributes, which is the meaning of
attribute reduction. Distinguishing ability of attribute sets can be well
described by relative knowledge granularity in domain. Therefore, how to use
the information based on relative knowledge granularity to simplify the
calculation of attribute reduction. It is an important direction of
research. For increasing productiveness and accuracy of attribute reduction,
in this paper we investigate attribute reduction method of relative
knowledge granularity in intuitionistic fuzzy ordered decision table(IFODT).
More precisely, we redefine the granularity of knowledge and the relative
knowledge granularity by ordered relation. And their relevant properties are
proved. On the premise that the decision results remain unchanged, in order
to accurately calculate the relative importance of any condition attributes
about the decision attribute sets, the conditional attribute of internal and
external significance are designed by relative knowledge granularity. And
some important properties of relative attribute significance are proved.
Therefore, we determine the importance of conditional attributes based on
the size of the relative attribute significance. In the aspect of
computation, the corresponding algorithm is designed and time complexity of
algorithm is calculated. Moreover, the attribute reduction model of relative
knowledge granularity of efficiency and accuracy is proved by test. Last, the
validity of algorithm is demonstrated by an case about IFODT.
Heuristic Attribute Reduction Based on Neighborhood Knowledge Granularity
