Research Of Multi-granulation Rough Sets Based On Tolerance Relation In Electronic Commerce Web Mining

Rough Set Theory (RST), proposed by Polish scholar Z. Pawlak in1982, is a powerfulmathematical tool which can be used to deal with vague, uncertain and inconsistencyproblems, In Pawlak’s rough set model, most of researches are in Complete InformationSystems (CIS). However, for various reasons systems delt by people are usually IncompleteInformation Systems (IIS) in actual problem. Therefore, the study of IIS from the viewpoints of rough set theory has become one of the important topics in research of RST. Butthe classical RST and a lot of extended rough set model are based on a single kind ofgranulation constructor, so Qian et al. from the view of granular computing, pointed out thatwe often need to describe concurrently a target concept through multi binary relations on theuniverse according to a user’s requirements or targets of problem solving. Therefore, theyproposed the concept of Multi-granulation Rough Set model.Tolerance relation is a very popular binary relation to deal with IIS. This dissertationfirst analyzes the extended rough set model based on tolerance relation, which includes thevariable precision rough set based on tolerance relation and Multi-granulation Rough Setbased on tolerance relation.This dissertation then integrates the good points of variable precision rough set andMulti-granulation Rough Set, proposes the variable precision Multi-granulation Rough Setbased on tolerance relation in the IIS, and discusses the properties of variable precisionmulti-granulation optimistic approximation and variable precision multi-granulationpessimistic approximation. Moreover, it proposes a heuristic algorithm of attribute reductionbased on attribute signigicance and discusses the acquisition of decision rules by incompletevariable precision Multi-granulation Rough Set based on tolerance relation.Finally, this dissertation applies incomplete variable precision Multi-granulation RoughSet based on tolerance relation into Electronic Commerce Web mining. By analyzingcustomers’ behavior and data in an Electronic Commerce site, and using the relevant theoryof the incomplete variable precision Multi-granulation Rough Set based on tolerancerelation, it gets customers’ behavior decision rules in the Electronic Commerce site. Theresult of experiments shows the correctness and effectiveness of the approach.