A General Deep Learning Model for Volatility Forecast Using Technical Features

This paper presents a deep learning-based approach aimed at developing a general model to predict volatility using the technical features of financial assets derived from trading prices. By focusing exclusively on technical features instead of fundamental data or other asset-specific features, the model can utilize data from a variety of assets, thereby taking advantage of deep learning architectures, which have superior performance when trained with large datasets. Utilizing a deep learning framework for forecasting future volatility offers improved results in numerous financial applications, such as optimizing investment portfolios, managing risk, and pricing derivatives. This study assesses the effectiveness of four different deep learning architectures in understanding technical contexts and compares these with conventional econometric methods, including the generalized autoregressive conditional heteroskedasticity (GARCH) model and linear regression applied to technical features. The findings demonstrate a promising methodology, even when relatively simple technical features are used. JEL Classification: C15 , G11