Research And Application Of BN Modeling Method Based On Small Data Set

Unmanned Aerial Vehicle(UAV)has repeatedly demonstrated its outstanding military superiority in the modern battlefield and has received extensive attention at home and abroad.Threat assessment is the core part of the UAV mission planning system,but in the actual battlefield,the amount of data that can be obtained is limited,which seriously affects the quality of threat assessment modeling of drones.Therefore,thesis uses Bayesian Network(BN)to study the modeling method under small data set conditions.The main research contents of thesis are as follows:(1)An improved bird swarm algorithm is proposed for the problem that the classical bird swarm algorithm is easy to fall into local optimum.Firstly,the adaptive inertia weight is introduced,and the search space is adaptively adjusted to make the algorithm easy to jump out of the local optimum.Secondly,the bird group is divided into producers and followers according to the fitness value,which accelerates the convergence speed of the algorithm.Finally,the effectiveness of the improved bird swarm algorithm is proved by numerical experiments on typical test functions.(2)This thesis combines the improved bird swarm algorithm with expert constraints for BN structure learning under small data set conditions,and proposes a BN structure learning algorithm based on improved bird swarm algorithm under small data set conditions.The information contained in the small data set is not comprehensive,and the expert constraints can make up for the lack of information and improve the initial network and search process of the improved bird group algorithm.In this thesis,under the condition of small data set,the BN structure obtained by the BN structure learning algorithm based on the algorithm,the classical bird swarm algorithm and the particle swarm algorithm is compared by using the Chest Clinic network,the Animal Characteristics network and the Car Diagnosis network.The simulation results show that the proposed algorithm is closer to the real structure under the condition of small data sets.(3)Taking the UAV’s autonomous reconnaissance air battle battlefield environment as the simulation background,based on the real-time observation data of the UAV and related expert domain knowledge,the algorithm is used to construct the UAV threatassessment model,and then the BN parameter learning is used.The maximum likelihood estimation algorithm performs parameter learning.Finally,the joint tree algorithm is applied to the inference calculation,which verifies the effectiveness of the modeling algorithm.In this thesis,the classical bird swarm algorithm is improved firstly,and the search efficiency of the algorithm is improved.Secondly,the improved bird swarm algorithm is applied to BN structure learning,and combined with expert constraints,the BN structure learning algorithm based on improved bird swarm algorithm under small data set conditions is proposed.Finally constructed the UAV threat assessment model to prove the effectiveness of the proposed algorithm.Figure 32,Table 16,and reference 74.