| Three-way concept analysis is the product of the combination of three-branch decision making and formal concept analysis.Compared with formal concept analysis,the greatest progress of this theory is that it can simultaneously study the information that "has" and "does not have" in formal background.Attribute clustering is a theory of aggregating attributes based on equivalence classes to generate new attribute sets.Attribute granulation is a kind of theory based on granularity tree and pruning to decompose attributes into sub-attributes and form a new attribute set.In the study of multi-granularity,the new formal background generated by attribute clustering,attribute granulation and other processes has an internal relationship with the original formal background,and the original three-way concept has an internal relationship with the new three-way concept obtained by attribute clustering.In the process of method selection and concrete implementation,the comparison of its effect on the three-way concept lattice is the main evaluation index.For example,the selection of equivalence classes and the adjustment of constraints have an effect on the three-way concept lattice after attribute clustering.And there are many pruning on the same grain size tree,so selecting pruning becomes the only way to determine the optimal direction of granulation effect.Under the background of three-way concept lattice,the research on attribute clustering and attribute granulation is still insufficient,and the efficiency cannot be measured by effective means,which seriously reduces the speed of concept differentiation and refinement,and requires a large number of redundant calculations.Taking this as the research background,this paper studies the characteristics of attribute clustering and attribute granulation,studies the difference between concept lattice and three-way concept lattices before and after attribute clustering and attribute granulation,and proposes methods and indicators to measure efficiency.The main innovations are as follows:(1)Through theoretical derivation,it is proved that there is a close internal relationship between the original three-way concept and the new three-way concept derived from attribute granulation,which is taken as the basis for measuring the efficiency of attribute granulation.The concept of pessimistic attribute clustering,optimistic attribute clustering and general attribute clustering is put forward to make the result of attribute clustering more practical.The relationship between general attribute clustering and other two kinds of attribute clustering is obtained.Furthermore,through the process of clustering and the formation of the three concepts,the difference between the original three-way concept and the new three-way concept obtained by attribute clustering is compared,and two constraint indices are proposed through the connection between the two concept.In this way,the method of quantifying the influence of attribute clustering on the three-way concept lattice is proposed.(2)Based on the relation of attribute granulation hierarchy,attribute granulation hierarchy is divided into attribute granulation hierarchy with partial order relation and attribute granulation hierarchy without partial order relation.Further,the definition of refinement coefficient is given,and the measurement function of refinement coefficient in two kinds of attribute granulation levels is elaborated respectively,so as to achieve the purpose of measuring the efficiency of different attribute granulation directions.In order to improve the practicability of this theory,a refinement coefficient acquisition algorithm is designed,and the effectiveness and superiority of the algorithm are evaluated. |
