On the Trade-Off between Multi-level Security Classification Accuracy and Training Time

Automatic security classification is a new research area about to emerge. It utilizes machine learning to assist humans in their manual classification. In this paper, we investigate the importance of the training time of the machine learner. To the best of our knowledge, this has not been analyzed in previous works. We compare various machine learning methods, including SVM, LASSO and the ensemble methods Adaboosting and Adabagging, with respect to their performance. The paper demonstrates that the computational cost of a method is an important part of its performance metric.