MULTI-FACTOR RISK MODELS FOR COMMODITIES IN A REGIME-SWITCHING ENVIRONMENT

Abstract

This thesis aims to provide an ex-post feature selection analysis to the commodities market in the timeframe between February 2020 and February 2022. It provides two feature selection models using LASSO regression for each of the major commodities under their low, medium, and high variance regimes – one using data across the entire time frame, the other with a 30-day rolling window. Both methods allow us to narrow down major factors particular to each regime of a commodity and help illustrate and explain macroeconomic trends among the drastically evolved commodities market during the COVID-19 era.