This work proposes a real-time network anomaly detection system using the integration of Snort, an open-source intrusion detection system, with various unsupervised machine learning methods. This research compares six various anomaly detection approaches using Local Outlier Factor, One-Class SVM, Isolation Forest, K-Means clustering, Gaussian Mixture Models, and Elliptic Envelope. This research implemented all six approaches and compare their performances. Also, input network traffic data extracted from Snort logs into our system, which processes the input data in real-time using a Streamlit dashboard user interface. A better detection accuracy is revealed using ensemble decision-making across various models according to experimental results, in which Isolation Forest and Local Outlier Factor report very high detection accuracy for network abnormalities. This research obtained low-latency response time appropriate for real-time intrusion detection with high accuracy and recall rates using the proposed system. This work is the first to contribute to the field of computer security through the proposed comparative study of various anomaly detection methods in a deployable implementation setup.