Research On The Method Of Ensemble And Knowledge Acquisition Based On Granular Concept Lattice

With the advent of Big Data era,the quantity of data is large and the growth is fast.Moreover the data show various features.Extracting valuable knowledge and information from a large amount of data becomes harder,at the same time,the demand of the processing in Multi-source information system and the distributed storage and parallel processing of data is increased.Concept lattice as a valid tool for data analysis can be used to dig implicit knowledge from data,and has gained many achievements in recent years.However,the construction complexity of concept lattice is the main restrictive factor in its application.The newest researches of concept lattice are attribute reduction,reducing the complexity of both concept lattice construction and extracting rules.In this thesis,we mainly study the granular concept lattice model and its construction method,as well as the method of ensemble and knowledge acquisition based on granular concept lattice,the main work is as follow.In the aspect of granular sub-concept lattice,we use an approximate relationship to form the background coving,then choose the granular sub-context to construct concept lattice,and do study on granular concept lattice.The properties of granular concept lattice is gave,the relationship between the granular sub-concept lattice and original concept lattice is proved,moreover we provide a method for the combination between two granular concept lattices and proved that the original concept lattice can be constructed by combination of granular concept lattices.Finally,the experiments verified the effectiveness of this method.We also did some research on the relationship between two rule sets extracted from granular concept lattice.In the aspect of extracting and combination on rules,we propose a new concept called pseudo rule.Users can flexibly generate rules by pseudo rules.Then we provide a method of the combination between two pseudo rule sets.The merged pseudo rule set is corresponding to the pseudo rule set extracted from the concept lattice obtained by two concept lattices merged.All rules can be generated by pseudo rules.Finally,the experiments verified the effectiveness and fesibility of this method.In this thesis,we study on the combination of concept lattice and granularity,and a fast method for extracting rules based on horizontal splitting context.We adopt the idea of granularity on the splitting context,so we can get the granular sub-concept lattice with special features,also we adopt the idea of distributed processing on the processing of granular concept lattice and rule extracting.The constraint of concept lattice is reduced,the pace of extracting rules is increased,the actual application of concept lattice is enhanced.However,there are many issues to be studied in the application of concept lattice and granular computing in the knowledge acquisition.