Time Series and Machine Learning Volatility Forecasting

Time Series and Machine Learning Volatility Forecasting

• Pascal Letourneau, University of Wisconsin-Whitewater
• Lars Stentoft, Western University, Canada

Abstract:

This study revisits forecasting of equity volatility in the context of option pricing and option trading. We find that selecting a best model and hyperparameters depends on the metric used. In particular, results show that for simple models, a compromise has to be made in the size of the estimation sample, whereas more complex models do not suffer from using larger estimation samples. Contrary to general beliefs, a very flexible GARCH specification does not suffer from overfitting and provides equal or better out-of-sample forecasts. We combine time series forecasting under ensemble bagged trees to further improve the forecasting quality. Finally, we show how improved volatility forecasting can be used in option trading strategies.

 

All Illinois Tech faculty, students, and staff are invited to attend.

The Friday Research Presentations series showcases ongoing academic research projects conducted by Stuart School of Business faculty and students, as well as guest presentations by Illinois Tech colleagues, business professionals, and faculty from other leading business schools.

Join session on Blackboard Collaborate