From Twitter to Swift: Building Anomaly Detection.
Aaron Beckley
Follow
Published in
Towards Data Science
12 min read
Just now
—
Twitter (now X), back in 2015 made an Anomaly Detection Algorithm for use in tracking trends among their millions of users.
GitHub – twitter/AnomalyDetection: Anomaly Detection with R
Anomaly Detection with R. Contribute to twitter/AnomalyDetection development by creating an account on GitHub.
github.com
This package, made entirely in R, is still very usable. It was designed to be able to detect global and local anomalies, and it is able to successfully detect a wide variety of anomalies. For a complete list of what it can and can’t detect please check out Anomaly.io’s test of the original algorithm, as it is very comprehensive.
Why not 🤷‍♂️? I was bored.
Twitter’s Anomaly Detection Algorithm is a statistical framework designed for detecting anomalies, or outliers, in a time-series dataset.
There are two main core components to the algorithm.
- Seasonal Decomposition: The algorithm…