Anomaly Detection for Monitoring

Table of contents

1. Foreword
2. 1. Introduction
   1. Why Anomaly Detection?
   2. The Many Kinds of Anomaly Detection
   3. Conclusions
3. 2. A Crash Course in Anomaly Detection
   1. A Real Example of Anomaly Detection
   2. What Is Anomaly Detection?
   3. What Is It Good for?
   4. How Can You Use Anomaly Detection?
   5. Conclusions
4. 3. Modeling and Predicting
   1. Statistical Process Control
      1. Basic Control Chart
      2. Moving Window Control Chart
      3. Exponentially Weighted Control Chart
      4. Window Functions
   2. More Advanced Time Series Modeling
   3. Predicting Time Series Data
   4. Evaluating Predictions
   5. Common Myths About Statistical Anomaly Detection
      1. The Data Doesn’t Need to Be Gaussian
      2. Sample Distribution Versus Population Distribution
   6. Conclusions
5. 4. Dealing with Trends and Seasonality
   1. Dealing with Trend
   2. Dealing with Seasonality
   3. Multiple Exponential Smoothing
   4. Potential Problems with Predicting Trend and Seasonality
   5. Fourier Transforms
   6. Conclusions
6. 5. Practical Anomaly Detection for Monitoring
   1. Is Anomaly Detection the Right Approach?
   2. Choosing a Metric
   3. The Sweet Spot
   4. A Worked Example
   5. Conclusions
7. 6. The Broader Landscape
   1. Shape Catalogs
   2. Mean Shift Analysis
   3. Clustering
   4. Non-Parametric Analysis
   5. Grubbs’ Test and ESD
   6. Machine Learning
   7. Ensembles and Consensus
   8. Filters to Control False Positives
   9. Tools
      1. Graphite and RRDTool
      2. Etsy’s Kale Stack
      3. R Packages
      4. Commercial and Cloud Tools
8. A. Appendix
   1. Code