Realised Volatility Forecasting: Machine Learning via Financial Word Embedding

TL;DR

This paper explores how news-based NLP embeddings can enhance stock volatility forecasts, showing improved predictive accuracy especially during high volatility periods.

Contribution

It introduces a simple NLP framework that integrates news embeddings with volatility models, demonstrating improved forecasting performance and interpretability.

Findings

News embeddings improve volatility prediction accuracy.

Combining news with benchmarks yields consistent gains.

News themes identified as most relevant for volatility.

Abstract

We examine whether news can improve realised volatility forecasting using a modern yet operationally simple NLP framework. News text is transformed into embedding-based representations, and forecasts are evaluated both as a standalone, news-only model and as a complement to standard realised volatility benchmarks. In out-of-sample tests on a cross-section of stocks, news contains useful predictive information, with stronger effects for stock-related content and during high volatility days. Combining the news-based signal with a leading benchmark yields consistent improvements in statistical performance and economically meaningful gains, while explainability analysis highlights the news themes most relevant for volatility.