ANALYSIS OF THE EFFECTIVENESS OF MACHINE LEARNING ALGORITHMS IN THE CLASSIFICATION OF MILITARY EQUIPMENT IMAGES

Abstract

The article analyses the effectiveness of machine learning algorithms in classifying images of military equipment.It examines linear regression, decision trees, support vector methods, and deep neural networks according to criteria of training speed, prediction accuracy, scalability, and computational requirements. It has been established that neural networks demonstrate the highest accuracy (96 %), but require significant resources and training time (600 seconds).The support vector method provides an accuracy of 88 % with moderate resource consumption, while random forests show an optimal balance between accuracy (85 %) and training speed (80 seconds). Key efficiency factors have been identified: data size, model complexity, computational resources, and input data quality. Experiments conducted demonstrate that the training time for neural networks and the support vector method increases linearly in proportion to the increase in the number of images. At the same time, random forests exhibit a nonlinear increase in training time, indicating their ability to handle large sets of pictures of military equipment more efficiently.Optimisation strategies have been proposed using distributed computing and parallel processing to enhance the performance of the algorithms. For deep neural networks, using multiple GPUs is recommended, which can significantly reduce the training time of the model on large image datasets. Prospects for creating hybrid models that combine various algorithmic approaches to improve efficiency and accuracy in classifying images of military equipment have been outlined. Data cleaning and normalisation procedures are also essential to ensure high-quality input information.