Upcoming GRBIO seminars

Online, si voleu assistir-hi, contacteu amb grbio@grbio.eu

Archetypal and Archetypoid Analysis for Synthetic Data Generation

Privacy and governance constraints often prevent sharing individual-level data in health and sports analytics. This seminar presents a synthetic data generation framework built on Archetypal Analysis (AA) and Archetypoid Analysis (ADA): observations are represented via a small set of prototypes and their mixing weights, then new weights are sampled and mapped back to the feature space to produce synthetic records, optionally enriched via residual resampling. We discuss model selection for the number of prototypes, robustness across data regimes, and evaluation of fidelity, utility, and disclosure risk, with a focus on basketball data augmentation.

Bioscketch

Liukuan Yu completed his undergraduate studies in Financial Mathematics at Xi’an Jiaotong-Liverpool University, an international institution jointly founded by Xi’an Jiaotong University and the University of Liverpool. Then he moved to the UK to study Health Data Science at the University of Manchester for his postgraduate studies, where he systematically learned modern statistical and data analysis techniques, solidifying his determination to pursue advanced studies in biostatistics. Liukuan is now a PhD student at GRBIO under the supervision of Jordi Cortés and Daniel Fernández. His research primarily focuses on clustering analysis, particularly innovative applications of the k-means algorithm in sports analytics. His goal is to provide precise and scientifically supported data insights for the sports industry through statistical modeling.