Semicovariances and Machine Learning Methods in Oil and Gold Futures Markets
28 Pages Posted: 24 Mar 2021
Date Written: February 18, 2021
We adopt the semicovariance decomposition method and machine-learning models to forecast the realized correlation and realized volatility of oil and gold futures markets. The general framework consists of three steps: data preprocessing, accumulating window cross-validation, and performance evaluation. Data preprocessing generates a time-series of realized covariance matrices and semicovariance matrices as model input. Given input data at each time step, forecasting models are fitted with cross-validation. Forecasting models are evaluated using out-of-sample accuracy and robustness tests. In predicting the risk and correlation of oil and gold, the semicovariance decomposition method can achieve higher accuracy and stable results toward the various periods of training data compared to the benchmark DCC-GARCH model. A combination of random forest and the semicovariance decomposition method can increase the out-of-sample accuracy. These results show a possibility that the semicovariance decomposition combined with machine-learning techniques can be efficient even in the commodity futures market.
Keywords: Realized Volatility, Forecasting, Machine Learning, Semicovariance decomposition
JEL Classification: C32, C53, Q02
Suggested Citation: Suggested Citation