Neighborhood Systems Based Rule Acquisition In Formal Decision Contexts

Formal concept analysis(FCA)is an important data mining tool and has been widely used in many fields.The FCA concept analysis is formal decision contexts.Attribute reduction of formal decision context is mainly to use the relationship between the two concept lattices generated by the condition and decision attributes to remove redundant condition attributes.Attribute reduction and rule acquisition in decision formal contexts can discover the hidden knowledge more effectively,so attribute reduction and rule acquisition are two key research directions of FCA.This thesis combines the theories of formal concept analysis and rough set to obtain the neighborhood information systems based on equivalence relation,strong neighborhood relation and weak neighborhood relation,and then studies the approaches of attribute reduction and rule acquisition in three special decision formal contexts based on neighborhood systems.The main results are as follows:1.In formal decision contexts based on equivalence relation,the granular consistent set and granular reduction are defined,and the attribute reduction algorithm is formulated.By using the inclusion degree of set-valued vectors,optimistic rule fusion method and pessimistic rule fusion method in consistent formal decision contexts are proposed.2.The formal decision contexts based on neighborhood system is studied.Two new kinds of consistent set and reduction in the formal consistent decision contexts are defined.The attribute reduction algorithm of Boolean inference are formulated,and two rule fusion methods are constructed.Based on the inclusion degree on poset,four kinds of consistent sets in the formal inconsistent decision contexts are further studied,and their judgement theorems are proved.The Boolean method for calculating the four types of reductions is established.