A Comparison of Traditional Models and Machine Learning Techniques for Volatility Forecasting

No Thumbnail Available
Authors
Yoldas Ayan, Begum
Issue Date
2023
Degree
MSc in Financial Analytics
Publisher
Dublin Business School
Rights
Abstract
The calculation of volatility plays an important role in many applications such as derivatives pricing, portfolio optimization, risk-based value modelling and risk management. This study aims investigate and compare the efficacy of different forecasting methods for financial market realized volatility, specifically focusing on the SPX and DAX Index for identified range of time. The methodological design incorporated both machine learning techniques, such as Support Vector Machines (SVM), Neural Networks, and Long Short-Term Memory (LSTM) networks, as well as traditional econometric models like GARCH. Procedures followed a structured evaluation process, which involved training the models on a designated dataset, followed by validation and testing phases to measure their forecasting accuracy. Based on the findings, any machine learning techniques offer a more efficacious approach for forecasting volatility in the financial markets, compared to the traditional volatility models such as ARCH&GARCH models.