Decomposition Models for Agricultural Commodity Price Time Series: A Comparative Research

This paper explores applying decomposition models to deconstruct agricultural commodity price time series. It compares data decomposition using empirical wavelet transform (EW), empirical modal decomposition (EMD), singular spectral analysis (SSA), and variational mode decomposition (VMD), which could also decompose time series into trends and detailed components. The analysis is based on daily data from the Chicago Board of Trade (CBOT) corn closing prices from January 1980 to August 2021, with 10 456 observations. It is concluded that all four techniques are able to reduce the impact of noise, as well as capture the overall trend and main fluctuations, however, for the selected dataset, the singular spectral analysis (SSA) showed a better signal-to-noise ratio.