THE ARTIFICIAL INTELLIGENCE METHODS APPLICATION FOR ENEMY’S GROUP TARGET OBJECTIVES CLUSTERIZATION AND CLASSIFICATION

Abstract

The article shows that artificial intelligence methods, namely neural networks, can be used to solve the problems of clustering and classification of enemy group targets.
To solve the problem of clustering enemy group targets, a neural network of the SELFORGMAP type (selforganized maps) may be chosen, which consists of a layer of neurons and can classify a data set of vectors with any dimension. As revealed by the analysis of training the neural network for clustering enemy group targets, the NN SOMGTs has the best accuracy with 5 neurons in the hidden layer. The clustering of enemy group targets showed that they can be divided into 5 clusters.
The convolutional neural network such as the CNN GTs may ivebe chosen to classify enemy group targets. The training of the CNN GTs was carried out using the Levenberg – Marquardt algorithm. Based on the training results, the neural network for classifying group target objects was trained, verified and tested with an accuracy of 100%, the overall accuracy of the neural network for classifying enemy group target objects was 100%. The results of analyzing the histogram of errors between the target values and the predicted values after training the neural network for classifying group targets show that the error value is 0.00022. Since the errors are close to zero, the trained neural network performs well in classifying group targets. Testing the performance of the CNN GTs on an arbitrary set of group target objects showed 100% coincidence of the classes of actual group target objectives obtained using the CNN GTs.
The direction for further research could be the creation of neural networks to solve the problem of stributing a heterogeneous swarm of striking unmanned aerial vehicles over the objectives of an unsteady heterogeneous group target.