Three-way Concept Learning And Application Of Interval Valued Decision System

Three-way decisions model is an important tool to deal with uncertain information,and its idea of three-divide and rule is more consistent with human cognitive law.In the era of big data development,with the increase of data types,interval-valued data with stronger descriptive ability has been widely concerned by scholars.In practical applications,the data will be noisy or missing due to the influence of measurement,interference or information transmission,and it will show a dynamic increasing trend over time.If these data are ignored or deleted,it is likely to lead to the loss of useful information and misjudgment of decision-making.In this study,three-way decisions are applied to interval valued,and the specific contents are as follows:(1)Combining the attribute reduction method with the three-way decisions,the attribute reduction method of decision cost minimization for interval valued information system is proposed.Firstly,the interval-valued three-way decisions model is established.In order to distinguish the different decision partition costs among the compatible classes of different objects in the interval-valued decision system,a comprehensive loss function is established to describe the difference information between objects.Then,according to The Bayesian decision rules and the pre-given initial cost function,an adaptive threshold pair calculation method is constructed,and an interval valued system minimization decision cost attribute reduction algorithm is proposed.Examples and UCI data experiments verify the rationality and effectiveness of the algorithm.(2)Three-way decision models and incremental rule acquisition algorithm for incomplete interval-valued decision system are proposed.Firstly,the quantization similarity tolerance relation of incomplete interval-valued data is defined,and the equivalence class of the threeway decisions model is replaced by the quantization similarity tolerance relation of incomplete interval-valued data.Secondly,the dynamic rules of object sets are analyzed from two levels to obtain the policy.Finally,a set of UCI data sets is used to verify the proposed algorithm.Experimental results show that the proposed algorithm can not only reduce the loss of partition error,but also achieve higher partition accuracy,and has great advantages in running time.