Test submission 1

Abstract

The development of information technology has driven the growth of email-based communica-tion in business environments, including at Inti Everspring Indonesia. However, the high volume of incoming emails increases the potential for spam messages that may disrupt work effectiveness and data security. This study develops a spam detection system on the company’s website by ap-plying the Support Vector Machine (SVM) algorithm. SVM was selected because of its ability to perform text classification efficiently. The dataset used in this research comes from the company’s internal emails, consisting of labeled spam and non-spam messages. Since the dataset is imbal-anced, an oversampling process was applied, followed by text preprocessing steps including case folding, tokenization, removal of stop words, symbols, numbers, and stemming. The model was then trained using the SVM algorithm, and its performance was evaluated using several metrics: accuracy, recall, precision, and F1-score. Based on the experiments, the SVM-based spam detec-tion model achieved 100% precision, 100% recall, and a 100% F1-score. To validate the reliabil-ity of the algorithm, SVM performance was compared with BERT and Naïve Bayes. BERT achieved 96% accuracy, and Naïve Bayes achieved 97% accuracy. These results indicate that SVM is capable of classifying messages accurately, and SVM outperforms both algorithms.