A1418
Title: Covid-19 and commodity markets: A hybrid approach to temporal and spatial clustering
Authors: Charalampos Agiropoulos - University of Piraeus (Greece)
James Chen - Michigan State University (United States) [presenting]
Abstract: The purpose is to build upon prior research that applied unsupervised machine learning to evaluate commodity markets, exploring spatial and temporal dynamics. The conventional ontology of commodity markets, which categorizes precious metals, base metals, agricultural goods, and energy resources, was solidified through advanced clustering methodologies, focusing on daily logarithmic returns and conditional volatility forecasts. Furthermore, temporal clustering has been adept at pinpointing significant periods in energy-centric commodity markets, highlighting market shifts associated with geopolitical events and economic disruptions and, notably, the unprecedented challenges posed by the COVID-19 outbreak. A novel approach is introduced by fusing the spatial clustering methods with the temporal strategies, resulting in a unique hybrid methodology. The aim is to discern co-movements among asset classes during shifts from standard to extraordinary market states. Central to this investigation is the exploration of critical periods illuminated by temporal clustering to understand how foundational spatial relationships among financial assets adapt during economic upheaval. Though the immediate application focuses on energy markets, with an emphasis on understanding asymmetries in price co-movements and the intertwined relationship of biofuel and agricultural commodities, the methodology holds potential for broader applications.