I’m happy to announce that my implementation of the isolation forest unsupervised outlier detection algorithm was open sourced today. It is used by the LinkedIn Anti-Abuse AI team to detect and prevent abusive activity on the world’s largest professional network.
You can read more about isolation forests, my Scala/Spark implementation, how it solves challenges unique to anti-abuse work, and many other potential uses for this algorithm on the LinkedIn Engineering Blog. The repository is available on GitHub and artifacts are available on Bintray. Artifacts are also published to JCenter.