Probabilistic Distance Clustering for Compositional Data

This research explores the application of Probabilistic Distance Clustering (PDC) to compositional data, specifically addressing the challenge of handling a bounded sample space in distance computations. It first reviews existing log-ratio transformations that map compositions into Euclidean space for clustering. Then, it adapts PDC to handle compositional data by using the additive ratio combined with the Box-Cox transformation. A simulation study will evaluate this method compared to the well-established isometric log-ratio transformation.