Intertemporal Clustering of Commodity Indices

This study uses sophisticated statistical models and clustering ap- proaches to explore the intricacies of the dynamics of commodity markets. The GJR-GARCH model is essential to this research because it can han- dle financial time series asymmetries and ”fat tails” with ease. The GJR- GARCH model was manually implemented and compared to the standard library version to assess performance differences. Each of the 15 key com- modities was independently calibrated using historical data. The application of hierarchical agglomerative clustering (HAC) and k-means to datasets produced from daily conditional volatilities and log returns was also investigated in this study. In contrast to the first, less definite classifications by k-means, HAC found more complex structures in the data, revealing distinct market behaviours. In order to resolve possible temporal misalignments in data sequences, a novel technique combined Dynamic Time Warping (DTW) with HAC. This method revealed patterns that conventional linear approaches would have interpret the other way and offered a more specific view on clustering. A better comprehension of the clusters was also made possible by the use of Hidden Markov Models (HMM), which examined the behaviour of the clusters over time and examined important ideas like sustainability, volatility, and return. The creation of matching hedging strategies that were customised to the unique characteristics and behaviours of each clus- ter was influenced by this thorough investigation. In summary, this study highlights the significance of employing suit- able clustering and hedging tactics in commodities markets. The results demonstrate the potential of incorporating advanced mathematical tech- niques to improve our comprehension of market dynamics by striking a balance between model complexity and interpretability.