Neural method improves volatility forecasts with time series data

Nonparametric GARCH: An Iterated Neural Estimator

Machine Learning

Summary

Predicting how variable something like financial market volatility will be is tricky, especially when volatility changes in complex ways. The authors introduce a new approach that uses neural networks to estimate volatility without assuming a fixed model shape. They test their method showing it can learn better patterns and give more accurate forecasts, even when the data behaves nonlinearly or changes over time. Their approach also performs well on real financial data, including prices from options and high-frequency trades.

What this means in practice

  • For quantitative finance teams: Improve volatility predictions for risk models by using neural estimators that adapt to nonlinear and time-varying market behavior.
  • For algorithmic trading developers: Incorporate option-price and high-frequency data into variance forecasts using the described neural approach for refined trade strategies.$Commercial implications: Enables creation of advanced trading and risk-management software that outperforms standard GARCH models in dynamic markets.

Authors

Ruizhi Deng, Guohao Shen, Ngai Hang Chan

Abstract

We study nonparametric estimation and forecasting for GARCH-type processes and propose an iterated neural estimator that alternates between estimating the recursion map and updating the unobserved conditional variances. Under mild conditions, we establish finite-sample bounds for the time-averaged in-sample variance-path error and show that the estimator is minimax rate-optimal up to logarithmic factors. Under stationarity and mixing conditions, we establish the same rate for recursion-map estimation and derive finite-sample bounds for out-of-sample variance forecasts and expected excess Gaussian quasi-likelihood (QLIKE) loss. The convergence rates depend on the true lag orders and remain unchanged when the fitted orders are overspecified. Simulation studies illustrate the effectiveness of our estimators under nonlinear and time-varying volatility dynamics. Real-data applications suggest that our neural methods perform competitively and can incorporate option-price and high-frequency information.