Deteksi Email Spam dan Non-Spam Berdasarkan Isi Konten Menggunakan Metode K Nearest Neighbor dan Support Vector Machine

Abstract

Facing many email problems that have the potential to harm others. This abused email is commonly known as spam email, where the email contains advertisements, scams, and even malware. This study uses the K-Nearest Neighbor method and the Support Vector Machine to detect spam and non-spam emails based on content. The best value of the K-Nearest Neighbor algorithm is determined using Euclidean Distance measurements. Support Vector Machine and K-Nearest Neighbor can classify and detect spam or non-spam emails. K-Nearest Neighbor uses Euclidean Distance calculations with values K = 1, 3, and 5. The evaluation results use a confusion matrix, which shows that the K-Nearest Neighbor method with a value of k=3 achieves an accuracy rate of 92%, a precision level of 91%, a recall of 100%, and an F1-score of 95%. On the other hand, the Support Vector Machine method achieves an accuracy value of 97%, a recall of 100%, and an F1-score of 98%. This makes the Support Vector Machine method superior to the K-Nearest Neighbor method in this study. In addition, the model built can also be used to predict spam and non-spam from the contents of new e-mail content.

Top of Form

Keywords: Confusion Matrix, Email, KNN, Spam, SVM