A machine learning approach to volatility forecasting

TL;DR

This paper evaluates the effectiveness of various machine learning models in forecasting realized volatility of Dow Jones stocks, demonstrating competitive performance and insights into predictor importance.

Contribution

It introduces a minimal-tuning ML approach for volatility forecasting that outperforms traditional HAR models and proposes a new measure of variable importance.

Findings

ML models outperform HAR models in volatility prediction.

Forecast accuracy improves at longer horizons due to ML's ability to capture persistence.

The proposed variable importance measure reveals consensus and disagreement on predictor rankings.

Abstract

We inspect how accurate machine learning (ML) is at forecasting realized variance of the Dow Jones Industrial Average index constituents. We compare several ML algorithms, including regularization, regression trees, and neural networks, to multiple Heterogeneous AutoRegressive (HAR) models. ML is implemented with minimal hyperparameter tuning. In spite of this, ML is competitive and beats the HAR lineage, even when the only predictors are the daily, weekly, and monthly lags of realized variance. The forecast gains are more pronounced at longer horizons. We attribute this to higher persistence in the ML models, which helps to approximate the long-memory of realized variance. ML also excels at locating incremental information about future volatility from additional predictors. Lastly, we propose a ML measure of variable importance based on accumulated local effects. This shows that while there is agreement about the most important predictors, there is disagreement on their ranking, helping to reconcile our results.