Feature selection is a process that belongs to dimensionality reduction and is concerned with distinguishing relevant, irrelevant, and redundant variables for a task (e.g., a classification task). It aims to support in the removal of irrelevant and redundant variables to allow more simple and less computationally expensive models to be trained – to support model interpretability, reduce complexity and computational cost, support the performance as well as the reproducibility of models.
My research focuses on the use of feature selection and ensemble feature selection in the context of classification (supervised learning). Applications are often focused on microarray data sets that have a relatively high number features but only few observations, and on stock market forecasting.
I earned my Doctor of Philosophy (PhD) degree in 2020 from LUT University, where my doctoral dissertation focused on heuristic similarity-based feature selection methods for classification. I currently work as an assistant professor at the department of computer science at Reykjavik University in Iceland. Prior to that, I worked as an assistant professor and post-doctoral researcher at LUT-University in the research groups in “Finance and Business Analytics” in the Business School.