Inttention Recognition Of Group Cooperative Behavior Based On Suopprot Vector Machine

With the development of science and technology,modern military battlefield has entered the era of electronic warfare.In the airspace,independent and single tactical operations,such as fighter aircraft and bombers,have been unable to meet the operational needs of the modern battlefield.In this way,the collaborative approach of flight and fleet formation is the main form of combat warfare in the various national airspace in the world today.In the extremely complex and uncertain air battlefield,how to analyze the real-time situation of the battlefield and determine the behavioral intention of threat group cooperation is of great significance to our timely decision-making.In the air space,due to its high mobility and space uncertainty,the communication between group and group,as well as the members among a group is carried by wireless electromagnetic waves.The transmission of the reconnaissance and command information of both sides depends on wireless data communication chain.Therefore,the real-time dynamic of the battlefield inevitably has a close relationship with communication magnetic field environment.In this paper,the data link configuration and communication behavior of the USA main military in active service are selected to be the research objects.The communication behavior is simulated to obtain the data under different operational intentions,then the support vector machine are utilized to establish the intention recognition model of group cooperative behavior,and the recognition accuracy is verified.The specific contents are described below:Firstly,communication behavior and simulation data are carried out.Under the different operational intentions,the warplanes dispatched by the enemy are also different.Accordingly,the data link and the communication station configured on different combat aircrafts are also different.The communication event script is simulated by analyzing the correlative information.Combining the script,the data under different operational intentions are generated according to the cooperative behavior and rules of wireless communication,which is the data base of the intention recognition.Because the simulation data is generated from a preset script of the correspondence event and different data types are contained,it can not be used directly in machine learning.Moreover,the dimensions of each characteristic of the data is different,so,normalization process is beneficial to machine learning.Pretreatment is a necessary process to establish identification model.Secondly,a support vector machine parameter optimization method based on Bayesian Optimization Algorithm(BOA_SVMcg)is proposed.The test accuracy of SVM is selected as the preferential rules,and the excellent solution set is obtained from the initial population according to the preferential rules.The criterion BIC and Hill Climbing algorithm are applied to build the network model which reflects the internal relation of the data,and the network parameters are obtained based on the simulated data set.The new individuals are sampled based on the model.Coding method for constructing a network model in the process of BOA is improved to mainly show excellence in operation efficiency.Finally,the simulation experiment combining three parts is carried out.In the first part,the traditional binary coding method is compared with the improved binary coding method on the base of UCI standard data sets.,and the advantages of the new method in improving operation rate is verified;in the second part,the SVM parameters optimization algorithm based on BOA is compared with GS and GA on the base of UCI standard data sets,and the advantages of shorting the operation time and improving the testing accuracy are verified;in the third part,BOA_SVMcg was used to identify the intention of group cooperative behavior,and obtained the recognition accuracy of 94.76%.