Analisis Accuracy, Specificity, dan Sensitivity Artificial Neural Network pada Fitur Time Based Traffic

Abstract

Sebagai algoritma klasifikasi, sangat penting untuk mengetahui bagaimana kinerja ANN dengan mengukur tingkat akurasi, spesifisitas, dan sensitivitas yang dimiliki oleh ANN. Pengukuran dilakukan melalui beberapa tahapan yaitu compose training/testing dataset, pre-process training/testing dataset, determine the neural network secure, train neural network, dan test neural network dengan menggunakan KDD dataset. Hasil dari pengukuran kinerja tersebut menunjukan bahwa tingkat ANN memiliki accuracy (ACC) yang sangat tinggi yaitu 99.7%, tingkat specificity (SPC) 97,4% dan tingkat true positive rate (TPR) atau sensitivity yang merupakan kemampuan mendeteksi secara tepat suatu jenis serangan yaitu 99.8%.

As a classification algorithm, it is very important to know how ANN performance in measuring levels of accuracy, specificity and sensitivity owned by ANN. Measurements conducted over several stages, namely compose training/testing datasets, pre-process training/testing datasets, determine the neural network secure, train neural network, and test neural network by using KDD dataset. The result of the performance measurement showed that ANN get high level of accuracy (ACC) which is 99.7%, specificity (SPC) is 97.4% and true positive rate (TPR) or a sensitivity which able to detect exactly a type of attack is 99.8%.