Anomaly Detection Algorithm for Server Health Monitoring

📌 Overview

This project implements an anomaly detection algorithm to identify failing servers in a network based on system health metrics.
The algorithm models the distribution of server features and flags unusual behavior as potential anomalies, enabling proactive detection of failures.

---

🛠️ Tools & Technologies

• Python, Jupyter Notebook
• NumPy / Pandas for data manipulation
• Matplotlib for visualization
• Gaussian distribution + anomaly detection algorithm

---

📂 Dataset

• Simulated dataset representing server performance metrics
• Features include latency (ms) and throughput (mb/s)
• Goal: detect unusual server activity that may indicate failure

---

🔍 Methodology

1. Feature Analysis
   • Extract latency and throughput as server health indicators
   • Standardize features for distribution fitting
2. Modeling with Gaussian Distribution
   • Estimate parameters μ (mean) and σ² (variance)
   • Fit Gaussian contours to normal server data
3. Threshold Selection
   • Cross-validation to find best epsilon (anomaly threshold)
   • Use F1-score to balance precision vs recall
4. Anomaly Detection
   • Instances outside the probability threshold flagged as anomalies
   • Visualize anomalies with Gaussian contours

---

📊 Results & Insights

• Best epsilon (threshold): 1.377229e-18
• Best F1-score (cross-validation): 0.615
• Anomalies flagged: 117 instances
• Algorithm successfully detected failing server behavior beyond normal distribution bounds

---

📈 Example Visuals

Gaussian contours of server health data with anomalies (red markers):

---

🚀 How to Run This Project

1. Clone this repository:

   git clone https://github.com/Josefxl/Anomaly_Detection_Algorithm_for_Server_Health_Monitoring.git