Anomaly detection helps us find data points which are inconsistent with the rest of a dataset. But what does this mean when our data points have many individual features? In other words, how can we tell whether a data point is an outlier when we have dozens of input columns? In this talk, you will learn about several techniques designed to solve this problem. We will understand the intuition and math behind these techniques, implement a simple outlier detector in Python incorporating these algorithms, and create a Streamlit app to host the detector.
You will learn:
- The complexities around multivariate outlier detection
- Gain an understanding of some of the algorithms used in multivariate detection
- Build a simple outlier detection API to serve multivariate scenarios