Research On Rough Sets Based Decision Making Method For Situation Awareness

With the continuous development of economy,politics and the international situation,the risks with respect of public emergencies,financial risks,cyber security,military struggles and other complex scenarios became higher for human being.Through situational perception,situation assessment and situation prediction,situation awareness can detect,deal with and predict such complex management scenarios as early as possible.However,the collected information from various aspects in the data-driven complex management scenarios often shown as uncertain,incomplete.rough,fuzzy and dynamic,which brings great challenges to intelligent information processing.Although rough sets theory and three-way decisions are effective analysis methods to deal with the incompleteness,roughness and uncertainty,the research on the application of the granular eomputing method to the situation awareness scenario is still in the exploration and initial stage.Therefore,in light of granular computing,this paper systematically studies the decision-making method of rough sets theory from the perspectives of situation perception,situation assessment and situation prediction,with the background of complex management scenarios.The detail research content and results are presented as follows.1.Test-cost sensitive based rough sets model and situational perception of UCAV air combat scenarios.In the data-driven situational awareness scenario analysis problem,the acquisition of various types of data is often not free,but requires a certain test costs.Considering the test cost sensitivity,a test-cost sensitive a quantitative indiscemibility relation based rough sets model is proposed,which has realized the generalization of classical rough sets model.Moreover,in attribute reduction process,the acquire knowledge and minimum test cost criteria have been considered together and proposed a novel attribute fitness function.After the algorithm testing on several public data sets,the test-cost sensitive rough sets model has been applied to the situational perception task of UCAV air battle scene.2.Parallel and Sequential based multigranulation awareness strategies for complex management scenario.Awareness strategy is the soul of situational awareness task.In the process of situational awareness of various complex management scenarios,the data faced by decision makers are obviously multi-souree,multi-perspective and multi-granularity.Multigranulation cognition is one of the important means of human intelligence cognition.On the basis of previous work,this paper has conducted further research on parallel and sequential granularity cognitive methods.On the one hand,based on the analysis of the shortcomings of optimistic and pessimistic multigranulation decision-theoretic rough sets,a flexible multigranulation decision-theoretic rough sets model is proposed.On the other hand,a dynamic sequential three-way classification method based on local reduction is constructed,which takes local reduction as the starting sequence of sequential information granules and global reduction as the terminating sequence.3.Hybrid cost sensitive information fusion model and situation assessment in multi-source information systems.In this paper,a hybrid cost sensitive multi-granularity rough set model is proposed based on the background of the cost sensitive multi-source information system in the scenano of war game inference,which considering the test cost and decision cost into information fusion model simultaneously.In terms of information granularity reduction,the information granularity reduction problem has been transformed into an optimization problem,and designing a new adaptive function based reduction to meet different practical requirements.In addition,the proposed model and method are validated and tested with several public data sets.Finally,a multi-source information system in an actual scenario of war game deduction is taken as an example,and discusses the role of mixed cost sensitive multi-granularity information fusion model in situation assessment.4.Multi-granularity situation prediction method for eomplex data types.As we all know,heterogeneous and other complex data widely exist in the process of situation prediction.How to deal with such complex data and extract useful information are keys to situation awareness.In this paper,a novel multi-granularity classification prediction method is proposed in light of the justifiable granularity criterion.In the model learning stage,based on rough set theory,this paper first decomposes justifiable granularity into binary relations and justifiable attribute subspace.Then,the attribute subspace which conforms to the justifiable granularity criterion is constructed by attribute reduction method.In the stage of label prediction,three-way division strategies are used for prediction.Finally,the proposed method has been applied to the scenario situation prediction task of wargame deduction.