Explainability of outlier detection methods

Overview

Outliers are anomalous observations in a data set that are "outside the norm" of what would be expected. Identifying outliers is an important part of exploratory data analysis and data analysis in general. It is often a challenging problem and calls for advanced methods and approaches, including machine learning-based tools. As methods become more and more complex, their explainability becomes more difficult and more important. This research project will look at all aspects of explainability and explore new approaches and methods.

Research activities

Research activities could include but are not limited to discussing and recommending explainability methods for outlier detection.  Testing can be done on publicly accessible datasets that are commonly used in outlier detection research.

Outcomes

It depends on the level of engagement, in any case the results will be commiserate with the degree pursued.

Skills and experience

A data-science, mathematics, or computer science background is recommended.

Scholarships

You may be eligible to apply for a research scholarship.

Explore our research scholarships

Keywords

• outlier detection
• explainability

Contact

Contact the supervisor for more information.