Authors: Konstantinos Gkillas, Maria Tantoula, Christina Diakaki

Title: Machine Learning Jumps Detection and Volatility Modeling and Forecasting applied in Energy Markets

Abstract

We develop a nonparametric approach for jumps detection to accurately model and forecast the volatility of the energy market. To estimate energy market related realized volatility and jumps detection based on machine learning algorithms, we use high frequency price data. Specifically, we apply the machine learning models to identify patterns for jumps detection as the information content of jumps in future volatility is an important area of research in the financial forecasting literature; especially, considering that jumps have a substantial impact on future realized volatility.