Performance Analysis and Applicability Study of Machine Learning Models for Effective Threat Assessment of Missiles in Air Defense Systems

In military conflict situations, the accuracy of identifying projectiles is a critical element of national security. An air defense system performs thereat assessment of detected projectiles after identifying them as friendly or hostile. To enhance the accuracy of the air defense system, the threat level of projectiles should be evaluated accurately and swiftly. However, as the types of projectiles have recently become more diverse and a large number of projectiles attack, processing data using the existing method consumes a lot of time and money. There is a limitation that the accuracy of the air defense system can sometimes decrease. In this paper, research is conducted to discover a suitable machine learning model that can be applied to the air defense system. Among various machine learning models, DNN, SVM, and XGBoost are adopted to classify the types of projectiles based on their unique trajectory and velocity data. Each machine learning model is trained and tested using data generated by the actual air defense system. In order to verify the applicability of machine learning models to the air defense system in the real world, the performance of the models was compared and analyzed based on their prediction accuracy, computation time, and memory usage.