Research On Sequential Three-way Decision Model

Three-way decision is a kind of granular computing method to deal with uncertain decision-making.It is a "three way and rule" model in line with human cognitive process.The main idea is: in the decision-making process,the whole is divided into three parts or three particles,and different decision-making behaviors or processing strategies are adopted for different parts or particles.Three-way decision has been widely used in knowledge discovery,data mining and pattern recognition.Compared with the traditional two-way decision,the key of the three-way decision is to introduce the delayed decision,so as to avoid the loss caused by the wrong decision.When making the delayed decision,in order to make the next accurate decision,it is necessary to further collect the state information of the decision,that is,to transform the knowledge of the decision object from coarse-grained to fine-grained.This kind of decision-making process from coarse-grained to fine-grained constitutes a sequential decision-making method,which can describe the dynamic progressive decision-making process used in many practical problems.For example,in medical diagnosis,it is still impossible to determine which disease the patient has after the preliminary examination.At this time,it is necessary to gradually check the patient through other means until the patient's condition is finally determined and the corresponding treatment plan is given.In view of the limitations of two kinds of sequential three-way decision models,two kinds of improved sequential threeway decision models are proposed in this paper:(1)In view of the existing sequential three-way decision model based on the dynamic update of attribute values,In the decision-making process,on the one hand,the relevance between the attribute values before and after the update is not considered,on the other hand,the feature extraction algorithm of attribute values based on standard deviation makes the classification effect of the model unsatisfactory.Therefore,this paper defines the attribute value relevance degree to describe the relationship between the attribute values before and after the update,and proposes an attribute value feature extraction algorithm based on Gini coefficient.On this basis,an improved sequential three-way decision model based on the dynamic update of attribute values is proposed.In addition,two kinds of classified abnormal cases are given.Finally,the validity of the proposed model is verified by experiments.(2)The existing temporal-spatial sequential three-way decision model,considering the time factor of data collection and the spatial factor of multi-level granularity structure,constructs the multi-granularity structure of temporal-spatial fusion.Through the fusion of binary relations,the temporal-spatial sequential three-way decision is realized.The fusion method based on binary relation belongs to the fusion of data layer.Through the fusion of decision results,this paper proposes an improved temporal-spatial hybrid sequential three-way decision model,and gives three fusion strategies of decision results,which are optimistic fusion strategy,pessimistic fusion strategy and variable fusion strategy.By comparing the accuracy and efficiency of the model of temporal-spatial sequential three-way decision model based on binary relation fusion and the model of temporal-spatial sequential three-way decision model based on the fusion of decision results,the validity of the model is verified.