Signal simplification in volatility forecasting does not improve neural predictions
Signal Simplification Is Not Predictive Simplification: Diagnosing Residual Neural Forecasting in Short-Horizon Volatility
Machine Learning
Summary
Forecasting short-term changes in financial market volatility can be challenging. The authors looked at a common approach where a statistical model first simplifies the data, and then a neural network tries to predict the leftover part. They found that this simplification makes the leftover data easier to model but does not help the neural network make better predictions. In fact, using only the neural network without simplification yielded better forecasting accuracy.
What this means in practice
- •For quantitative trading teams: Avoid using residual neural network models in short-horizon volatility predictions to reduce runtime without losing accuracy.
- •For financial risk modelers: Use pure neural network models rather than hybrid statistical-neural pipelines for more accurate short-term volatility forecasts.
Authors
Bingqi Lian, Linfeng Cheng, Mei Lu, Jerry Wu
Abstract
Hybrid statistical-neural pipelines often assume that a successful statistical first stage leaves a cleaner and more learnable residual target. We examine that assumption in short-horizon volatility forecasting through a signal-forecast-system diagnostic framework. Across five liquid U.S. assets, a volatility-aligned HAR-style model outperforms AR, MA, and ARIMA. Within expanding training windows, the pre-standardization fitted residual process used to construct residual-LSTM sequences has about 82% lower variance than the corresponding target and near-zero lag-1 autocorrelation; independently, rolling pseudo-out-of-sample HAR errors show about 74% variance reduction and similarly weak lag-1 dependence. Residual-only LSTM augmentation nevertheless raises mean squared error from 0.3049 to 0.3594 on average, with deterioration on every asset. Pure LSTM records the lowest selected pseudo-out-of-sample MSE, 0.2649, while the residual hybrid requires substantially more end-to-end runtime without improving accuracy. We describe this pattern as forecaster-preconditioner asymmetry: first-stage forecasting success and statistical residual simplification need not translate into useful downstream neural preconditioning.