The Extention And Research Of Panweighted Multi-granularity Rough Sets Model

Purpose —— Based on the multi-granular rough set model,the Pansystems series-parallel rough set model is proposed by using the Panoperators to transform arbitrary relations into equivalence relations.However,when making decision,the Pansystems series-parallel rough set model has three defects: In the Pansystems disjunctive parallel rough set model,lower approximation conditions are too loose,and in the Pansystems conjunction parallel rough set model,lower approximation conditions are too strict;Without considering the inequality among the granularity spaces;It's liable to be affected by noise datas.In order to solve the above problems,the Pansystems series-parallel rough set model is further extended.Methods —— From three angles(the quantity,quality,fault tolerance),the Pansystems seriesparallel rough set model is extended;and combining the extended model with fuzzy clustering algorithm to solve the problem of boundary fuzzy.Finding —— Firstly,based on the number of granularity,the concept of variable precision is introduced,and the number of granularity is controlled by the precision parameters ? in the lower approximation set.So the Variable precision Pansystems series-parallel rough set model is put forward.Secondly,based on the inequality among the granularity spaces,the concept of Panweighted is introduced,and the Panweighted Variable precision Pansystems series-parallel rough set model is put forward.In this model,the importance degree of each granularity spaces is described objectively and quantitatively by Panweighted.Then,based on the fault tolerance,the idea of probability is introduced.The lower and upper approximation sets are defined through the conditional probability.So the Panweighted Variable precision Pansystems series-parallel probabilistic rough set model is put forward.Finally,an improved fuzzy clustering algorithm is presented,which combines the Panweighted Variable precision Pansystems series-parallel probabilistic rough set model and fuzzy clustering algorithm.It uses the upper and lower approximation sets to represent the clustering results.Through comparing the proposed algorithm with the K-mean algorithm and the fuzzy C-mean algorithm,it is proved that the algorithm can solve the problem of boundary fuzzy effectively and improve the clustering accuracyLimitations —— For the model parameters,combining with the Bayesian decision and loss function method to obtain objectively and reasonably;For the reduction algorithm,combining with the heuristic algorithm to obtain a more efficient algorithm.Practical implications---By extending the Pansystems series-parallel rough set model which can adapt to the distributed information system and complex system better,it will effectively and rationally to make decision analysis and rule extraction.Originality —— By introducing parameter ? to control the number of approximate size space;information entropy and the Pansystems are introduced to assessment the importance of the granularity space;by combined with the statistical probability,which makes granularity space havethe fault tolerance to noise data;an improved fuzzy clustering algorithm is presented,which combines the Panweighted Variable precision Pansystems series-parallel probabilistic rough set model and fuzzy clustering algorithm,through introducing the concept of upper and lower approximation to solve the boundary problem of fuzzy clustering results.