In this talk I will describe and motivate two different ensemble techniques which can be seen as extensions of the incredibly popular Random Forests, which I approach purely out of curiosity. The first is the 'Isolation Forest', which uses Random Forest for anomaly detection. The second is 'Random Jungles' which are an extension of Random Forests to ensembles of Directed Acyclic Graphs (DAGs) rather than simple trees. I'll talk about my experience in implementing and testing out these ideas, and why one might use them.
Random Jungles and Isolation Forests
Padarn recently moved to London after graduating from the Australian National University with a degree in Computational Mathematics. His interest is in the analysis, design and implementation of algorithms for problems involving uncertainty (so Machine Learning). He is currently a Data Scientist working at Blue Yonder in West London, working on predictive analytics.