Prediction Research On Plan Recognition In Military Confrontation

In the half century since the birth of artificial intelligence,the changes brought to the field of computer science can be described as earth-shaking.As a branch of artificial intelligence,planning recognition has also made great progress and produced many research results.The continuous update of technology and results provides people with more convenient and faster ways to live and work.In the early stage of development,many fields are used in planning recognition,mostly focusing on natural language understanding,intelligent user interface,and user models.Related areas.So far,it has developed into the fields of network security,intrusion detection,and military tactical planning.With the continuous update and development of artificial intelligence and planning recognition theory and technology,the continuous emergence of high-tech militarized equipment will inevitably have a major impact on the battlefield situation and the key to victory.This means that the battlefield has undergone major combat forms.change.Therefore,it is of great significance to apply planning recognition to the military field.The article mainly conducts research from the following aspects:Firstly,in view of the relatively weak capabilities of general plan recognition algorithms,an improved plan recognition method based on Ada Boost is proposed.Regarding each plan recognition prediction model as a weak predictor,using the core idea of the Ada Boost algorithm,multiple weak predictors with small errors after multiple training are combined into a strong predictor.Use the obtained strong predictor to identify and predict hostile plans in military confrontation to improve the recognition accuracy of the plan recognition algorithm.Second,the article proposes a multi-layer plan recognition method under the condition of multiagents—MPRMC.This method is mainly aimed at hostile behaviors with more complex data structures.Before prediction,the bag of words model and TF-IDF model are used to extract the features of the data.This method pre-sets multi-layer Agents to collaborate in parallel,aiming at the original complex data input The features are abstracted layer by layer,that is,the feature attributes extracted from each layer are used as the input of the next layer.While extracting data features,an activation function is added to the output of each layer,so as to avoid the adverse effects of linear connection and ensure the recognition effect.Finally,the algorithm is tested with the more complex aerial bombing operations data set,and the experimental results prove that the algorithm has a better recognition and prediction effect.Third,the two algorithms proposed in the article are detected in the field of hostile behavior.In the experiment,another feature extraction model-denoising autoencoder is used.Then,two comparative experiments were carried out on the basis of the two algorithms proposed in this thesis.The first experiment is to ensure that the core algorithm remains unchanged.On this basis,the noise-reducing autoencoder and three feature extraction models proposed by other researchers are applied.The experimental results prove that it is desirable to use the noise-reducing autoencoder to extract the features of hostile behavior.The second experiment is to ensure that the feature extraction model remains unchanged.On this basis,the two algorithms in this thesis are compared with the three algorithms proposed by other researchers.The results prove that the two algorithms in this thesis have good recognition and prediction accuracy.