Less is more: Feature selection in data analysis

Abstract: In this talk, we will delve into the critical role of feature selection in managing big data. We will discuss the challenges posed by large datasets and the need for preprocessing to ensure data quality and learning efficiency. The presentation highlights dimensionality reduction techniques, distinguishing between feature extraction and feature selection—transforming or selecting features to enhance model performance. We will also explore different feature selection methodologies, including filters, wrappers, and embedded methods, emphasizing their impact on model transparency and interpretability. The talk concludes by addressing future challenges in feature selection, particularly in deep learning and data bias, and the need for scalable methods.

Bio: Verónica Bolón-Canedo received her B.S. (2009), M.S. (2010) and Ph.D. (2014) degrees in Computer Science from the University of A Coruña (Spain). After a postdoctoral fellowship in the University of Manchester, UK (2015), she is currently an Associate Professor in the Department of Computer Science and Information Technologies of the University of A Coruña. She received the Best Thesis Proposal Award (2011) and the Best Spanish Thesis in Artificial Intelligence Award (2014) from the Spanish Association of Artificial Intelligence (AEPIA). She has extensively published in the area of machine learning and feature selection. On these topics, she has co-authored two books, seven book chapters, and more than 100 research papers in international conferences and journals. She co-organized several special sessions at international conferences, such as ESANN and IJCNN, and served in program and scientific committees. Her current research interests include machine learning, feature selection and green algorithms. She has served as Secretary of the Spanish Association of Artificial Intelligence and is member of the Young Academy of Spain and the Royal Academy of Sciences of Spain.