The Research Of A Matching Method On Knowledge-based Battlefield Data Sample Label

Multi-service cooperative operation has become the main form of information warfare nowadays.In the face of complex combat system,the comprehensive and accurate understand of commanders in battlefield situation is the basis for making correct decisions and occupies an important position in situation assessment.As the extraction and generalization of battlefield sample data,battlefield data sample label carries important information of battlefield sample data.Therefore,the research on battlefield data label matching plays an increasingly important role in the field of battlefield situation recognition.In order to achieve the goal of label matching,a standardized and reasonable label framework is established for battlefield data samples,a numerical definition of battlefield data sample labels is given and the Gradient Boosting Decision Tree(GBDT)algorithm is used to realize the label matching of battlefield data samples.Firstly,the main work of domestic and foreign scholars is introduced in the field of label matching and situation recognition,the research progress in the field of ontology and the basic principle and theory of regression algorithm in machine learning are also mentionedSecondly,according to the characteristics and characteristics of the actual battlefield,a framework of ontology-based situation knowledge base is designed,including the construction of situation ontology and ontology-based situation knowledge base.The construction process of situation ontology includes describing the abstract concept of battlefield situation elements and determining the method of building ontology.Ontology-based knowledge base construction process includes three parts:defining the classification of situation concept,explaining the attributes of situation concept,and constructing situation rule base with logic and production representation method.In this paper,four potential label frameworks are constructed for tasks,resources,capabilities and battle situations,and their functions and meanings are explained in detail.Among them,mission potential label describes the battle mission information qualitatively,resource potential label and capability potential label describe the battlefield resources and combat capability quantitatively,and battle situation label describes the overall battlefield situation.In view of the characteristics of battlefield data sample labels,such as many information and strong correlation,GBDT algorithm is chosen to realize label matching process.For the unknown battlefield data samples,the preliminary data screening is carried out,and the GBDT multi-classification algorithm is used to match the task potential labels;for the resource potential labels and the capability potential labels,because of continuous real values,the GBDT algorithm is used to regression to obtain the matching results,and the battlefield situation labels are added manually to complete the matching of the four situation labels.The matching model of battlefield data sample label is designed.Finally,according to the above methods and models,the label matching verification experiments using the data of the wargame deduction platform are carried out.For task and capability labels,the accuracy and proximity of label matching an d actual label results are calculated respectively.Similar labels in knowledge base are added automatically according to label matching results.Finally,by displaying all labels in the corresponding battlefield data samples,it proves that proposed matching model based on knowledge and GBDT algorithm is feasible and practical.