Research On Three-way Risk Decision Model And Its Applications

As a kind of new granular computing methods,the three-way decision model which can extract useful knowledge from uncertain information reasonably and effectively is more in line with human cognitive.As the origin and main research branch of the threeway decision theory,the three-way decision model with decision-theoretic rough sets has become an important tool for dealing with uncertain information.However,there are still several problems in the three-way decision model with decision-theoretic rough sets,which deserve further study.For example,in the current study,when constructing risk measure functions,the difference information among equivalence classes is ignored.Thus,this thesis deeply studies the three-way decision model with decision-theoretic rough sets and its improved models.Then,a novel three-way decision model with decision-theoretic rough sets using utility function is proposed,which performs better.Then,an adaptive three-way classification model based on similarity is proposed and applied to the spam filtering to meet the high-precision requirements.Concretely,the main work of this thesis is as follows:(1)Based on the Euclidean distance,a function for measuring the difference between the equivalence class and the target set is given to explain that there is difference information among the equivalence classes.Based on utility function,a function to score condition attribute values is given,and the properties of the scoring function under different strategies are presented.And,a method for calculating attribute weight is given based on knowledge distance.Then the utility-based risk measure function is constructed,and its properties of the function are discussed.Under Bayesian decision rules,a utility-based three-way decision model is presented.The experimental results validate the superiority of the proposed model in performance.(2)In order to improve the classification accuracy of the spam filters,based on the the previous work,an evaluation function for condition attribute values is proposed based on similarity.According to the variance of condition attribute range,a method for calculating the weights of condition attributes is given.By weighting the evaluation values of each condition attribute value,a comprehensive evaluation function is constructed,and the related properties are discussed.Based on Bayesian decision rules,an adaptive three-way classification model is proposed.And,the experimental results show that the proposed model has a good effect for spam filtering.