I have been working on Outlier Detection(OD) for 4 months now and what amazed me the most is that a complete and scalable OD package dedicated for python is lacking. Fortunately, pyOD has just been released to fill the gap. Over 20 OD algorithms which could benefit my work and research in AIOps have been meticulously implemented in the best python OD package ever coded.
pyOD
An brief introduction to pyOD can be found here. It’s a brief introduction overall but quickly getting a grasp of not only the algorithms but also how the package works in general is of great importance.
Paper with code
Aha, here is where I first found the pyOD package. Two functionalities that make it a great website for a data scientist or machine learning engineer to visit on a regular basis:
- Provide with code repositories for each ML paper;
- Showcase the state-of-the-art datasets/algorithms/papers in each SIG or sub-SIG.
However, an algorithm being state-of-the-art doesn’t mean that it will necessarily transfer to better accuracy and precision in real-life detection. Prototyping and testing are inevitable en-route to the holy grail.