Forecasting white maize futures volatility through hybrid GARCH-based deep learning models
We would like to invite all enthusiasts and researchers of the financial world to the next Koźmiński University Finance Research Seminar Series event!
When? June 29th 2022 at 12:15 PM
Where? MS Teams Access (remote access)
Details about remote access will be sent to the provided e-mail address once the registration is completed. The registration form is available below.
Abstract Grain price volatility has significant socio-economic implications, heightened in the context of food security and renewable energy. Grains such as maize serve as a staple food for low-income households, predominantly in emerging economies like South Africa. Farmers in these economies are tasked with production decisions that often need to be made before prices are known. Hence, the ability to accurately forecast grain prices therefore not only protects the profits of farmers, but also filters down to the prices consumers pay for food. This paper employs several hybrid deep learning models to increase the volatility forecasting precision of white maize futures prices. Key input variables include SAFEX-listed yellow maize, wheat, and sunflower seeds; CBOT-listed corn and the Rand/Dollar exchange rate. Forecasting improvements over traditional models are demonstrated through a heteroscedasticity-adjusted mean squared error (HMSE) measure.
Keywords: white maize volatility, hybrid deep learning, GARCH models, heteroscedasticity-adjusted mean squared error.
Special guest and speaker: Dr Chun-Sung Huang
Dr Chun-Sung Huang is an Associate Professor at the University of Cape Town, and currently heads the Finance Postgraduate Study programmes in the Department of Finance and Tax. Chun-Sung holds a PhD in Financial Mathematics from the University of Cape Town. His research interests lie in both computational finance and financial risk management. Specifically, Chun-Sung has a keen interest in the valuation of exotic-style derivatives through numerical integration techniques, as well as the estimation of volatility, Value-at-Risk and Expected Shortfall through novel quantitative methods. More recently, he has developed a keen interest for the implementation of deep learning models to tackle some of the challenges embedded in the above-mentioned.