On the Classification of Intermediate Range Missiles During Launch

Abstract

This thesis explores the comparisons and contrasts of different statistical method approaches to a missile system classification which is sufficiently rapid to be suitable for use during engagement. It is meant to extend the work published by Eckert et al. by using simulated noise from common radar equations and running the dataset through an α-β filter. Additionally, this work expands on their work with the inclusion of another supervised learning technique, the support vector machine. The ultimate goal of this series of work is the rapid quantitative and statistically defensible descriptions of unknown missiles using the simulated telemetry data scheme during flight. The primary two supervised statistical learning methods explored in this thesis are Artificial Neural Networks (ANN) and Support Vector Machines (SVM). Of these two different statistical learning techniques, SVM classifiers were found to be the optimal classification technique in the context of this framework; SVMs were able to correctly identify 100\% of the testing dataset and guarantee a global optima.