<![CDATA[Network attacks pose a serious threat to the security of digital infrastructure, making accurate and efficient detection an urgent necessity. The National Cyber and Crypto Agency (BSSN) recorded 330,527,636 instances of anomalous traffic or network attacks. This sheer volume renders manual handling by cybersecurity personnel highly impractical. Consequently, a detection system based on Machine Learning (ML) is required. However, single machin e learning models have proven suboptimal in effectively resolving this complex problem. This research aims to develop a robust Network Intrusion Detection Model by employing advanced Ensemble Learning techniques—specifically Bagging, Boosting, and Stacking—to significantly optimize the performance of the base ML models. The study utilizes the UNR-IDD dataset. The methodology begins with comprehensive Data Preprocessing, including Min-Max Scaling and Dimensionality Reduction. Model performance is comprehensively evaluated using classification metrics across three distinct data splitting scenarios (70:30, 80:20, and 90:10) to identif y the optimal configuration. The experimental results demonstrate that the Stacking Ensemble Learning approach achieves the most optimal accuracy among the tested methods.Keywords: Network Intrusion Detection; Machine Learning; Ensemble Learning; Stacking; Dimensionality ReductionAbstrakSerangan jaringan telah menjadi ancaman serius bagi keamanan inf rastruktur digital . Badan Siber dan Sandi Negara (BSSN) mencatat terdapat 330.527.636 kasus traff ic anomali atau serangan jaringan. Hal tersebut sangat tidak memungkinkan personal keamanan siber untuk menanganinya. Maka, dibutuhkan sistem deteksi berbasis Machine Learning . model Machine Learning tunggal belum optimal dalam deteksi serangan jaringan. Penelitian ini bertujuan untuk memodelkan deteksi serangan jaringan menggunakan Machine Learning dengan teknik ensemble learning bagging, boosting, dan stacking, guna untuk mengoptimalkan perf orma model Machine Learning. Penelitian ini menggunakan dataset UNR-IDD. Penelitian ini bertujuan mengembangkan Model Deteksi Serangan Jaringan yang robust menggunakan teknik Ensemble Learning. Proses dimulai dengan Data Preprocessing, meliputi Min-Max Scaling dan Reduksi Dimensi. Model dievaluasi komprehensif menggunakan metrik klasif ikasi. Pada tiga skenario pembagian data (70:30, 80:20, 90:10) untuk mengidentif ikasi konf igurasi optimal. Hasil dari penelitian ini menunjukan bahwa Ensemble Learning stacking memilii akurasi yang paling optimal, dengan keseimbangan akurasi, presisi , F1 Score, dan Recall mencapai 100% . ]]>
