Volatility-aware LSTM improves daily stock price forecasting accuracy
Learning volatility-adaptive forgetting via GARCH-modulated LSTM for stock price forecasting
Computational Engineering, Finance, and ScienceMachine Learning
Summary
Stock prices change in unpredictable ways, sometimes being very volatile and other times more stable. Traditional forecasting methods using LSTM neural networks don’t adjust how they remember past information based on this volatility. The authors created a special LSTM that modifies its memory forgetting step using estimates of market volatility, making it remember recent data more during turbulent times. Their experiments show this approach predicts stock prices more accurately and stably than existing methods.
What this means in practice
- •For quantitative traders: Implement improved intraday trading algorithms that adjust memory retention based on market volatility for more accurate price predictions.
- •For financial data engineers: Ingest volatility estimates directly into neural network gates to enhance forecasting pipelines for stock price time series.
Authors
Xin-Rui Lu, Da-Qing Zhang
Abstract
Stock price forecasting is hindered by the heteroscedastic nature of financial time series, in which turbulent and tranquil regimes alternate. Although the Long Short-Term Memory (LSTM) network is widely used for this task, its gates are oblivious to such volatility variation, so memory retention remains fixed regardless of market conditions. A common remedy is to supply volatility estimates as an additional input feature, yet this leaves the internal memory dynamics untouched. Instead, volatility can serve as a direct regulatory signal over the gate activation itself. This paper presents the Gsigmoid-LSTM model, which embeds Generalized Autoregressive Conditional Heteroscedasticity (GARCH(1,1))-derived volatility estimates into the LSTM forgetting gate through the Gsigmoid activation function, a volatility-adaptive variant of the sigmoid that combines a volatility adjustment factor with slope regulation via a learnable parameter. The model thereby automatically prioritizes recent information during turbulent periods while preserving long-term dependencies in stable markets. In the engineering application of daily stock price forecasting, extensive experiments, including activation function comparisons, baseline evaluations, and ablation studies, demonstrate that this structural gating modification effectively improves forecasting accuracy and robustness. The proposed model significantly outperforms a feature-level fusion baseline carrying the same volatility signal on all five datasets, as confirmed by Diebold-Mariano tests with all p<0.01, attains the best average rank of 1.00 in the Friedman test across nine models, and achieves the lowest MSE and MAE on all five datasets.