Research On Multiview Sequential Three-way Decisions

Three-way decision is a ‘three way and rule' model when processing uncertain information.Unlike the two-way decision that only provides accept and refuse decision strategies,three-way decision introduces the choice of delayed decision to avoid the possible costs or losses from wrong decisions.As a typical granular computing model,it is closely related to the granular structures on problem description and solving.From the perspective of granular computing,multiview and multilevel are two basic principles of problem solving.Existing three-way decision models describe and solve problems using either the multiview granular structure or the multilevel structure,they are classical three-way decision,sequential three-way decisions,multigranulation three-way decisions,and multigranulation sequential three-way decisions.The classical three-way decision is based on the one view and one level granular structure;the sequential three-way decisions is based on the one-view and multilevel granular structure;the multigranulation three-way decisions is based on the multiview and one level granular structure;and the multigranulation sequential decision is based on the multiview and one level granular structure,with the sequential process realized based on varied threshold values.Partition order product space as a new granular computing model solves and describes problems from multiview and multilevel,and follows the multiview and multilevel principles.This paper studies the three-way decision models based on the partition order product space.The main research content includes:(1)The multiview sequential three-way decision models is proposed based on the partition order product space.The depth-first and breadth-first search algorithms are proposed to identify the problem solving level in the partition order product space.The optimistic and pessimistic integration strategies are introduced to solve the problems.The search algorithms and integration strategies are combined to come up with four multiview sequential three-way decision models,namely the depth-first optimistic multiview sequential three-way decisions,the depth-first pessimistic multiview sequential three-way decisions,the breadth-first optimistic multiview sequential three-way decisions,and the breadth-first pessimistic multiview sequential three-way decisions.The models solve problems at multilevel from multiview.Existing three-way decision models may be considered as special cases of the proposed models.Experimental results demonstrate the effectiveness of the proposed models.(2)The multiview sequential three-way decision models based on the partition order product space,it uses a division in the universe to define a level,and uses multilevel with linear order relationships to define a view.The Cartesian products of multiple linear order relations form the partition order product space,which is a lattice.When there are redundant multiview and multilevel,the partition order product space would be huge and complicated,causing great costs and low efficiency when solving the problems with the multiview sequential three-way decision model.Based on the heuristic attribute reduction algorithm,the view significance and level significance are defined,generating the view reduction algorithm and the level reduction algorithm.The view and levels are thus reduced to simplify the partition order product space.Experimental results demonstrate the effectiveness of the proposed algorithms.