Paper: arXiv 2410.00288
Authors: Zeda Xu, John Liechty, Sebastian Benthall, Nicholas Skar-Gislinge, Christopher McComb
Abstract
Volatility, which indicates the dispersion of returns, is a crucial measure of risk and is hence used extensively for pricing and discriminating between different financial investments. As a result, accurate volatility prediction receives extensive attention. The Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model and its succeeding variants are well established models for stock volatility forecasting. More recently, deep learning models have gained popularity in volatility prediction as they demonstrated promising accuracy in certain time series prediction tasks. Inspired by Physics-Informed Neural Networks (PINN), we constructed a new, hybrid Deep Learning model that combines the strengths of GARCH with the flexibility of a Long Short-Term Memory (LSTM) Deep Neural Network (DNN), thus capturing and forecasting market volatility more accurately than either class of models are capable of on their own. We refer to this novel model as a GARCH-Informed Neural Network (GINN). When compared to other time series models, GINN showed superior out-of-sample prediction performance in terms of the Coefficient of Determination (\(R^2\)), Mean Squared Error (MSE), and Mean Absolute Error (MAE).
Complexity vs Empirical Score
- Math Complexity: 7.0/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical concepts such as GARCH model formulations and deep learning architectures with LSTM, indicating high mathematical density. It demonstrates strong empirical rigor through rigorous backtesting on multiple global indices, using standard error metrics like R2, MSE, and MAE, and provides detailed methodology for implementation.
Research Flowchart
flowchart TD A["Research Goal<br>Predict Financial Volatility<br>More Accurately"] --> B["Key Inputs<br>Time Series Data"] B --> C["GARCH-Informed Neural Network (GINN)<br>Hybrid Methodology"] C --> D["GARCH Module<br>Captures Stylized Facts"] C --> E["LSTM Module<br>Deep Learning Flexibility"] D --> F["Computation<br>Physics-Informed Training"] E --> F F --> G["Key Outcomes<br>Superior Forecast Accuracy"] G --> H["Metrics<br>R² ↑ | MSE ↓ | MAE ↓"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a hybrid deep learning model called GARCH-Informed Neural Network (GINN) that combines GARCH with an LSTM network for volatility forecasting.
Source sentence
we constructed a new, hybrid Deep Learning model that combines the strengths of GARCH with the flexibility of a Long Short-Term Memory (LSTM) Deep Neural Network (DNN), thus capturing and forecasting market volatility more accurately than either class of models are capable of on their own. We refer to this novel model
Assumes 2
- The weight term λ in the GINN model lies between 0 and 1 and is selected via out‑of‑sample validation. model
Source sentence
Different values of λ were tested ranging from 0 to 1, with 0.01 increments between 0 and 0.2, and 0.05 increments between 0.2 and 1.
- Training converges after 300 epochs for both LSTM and GINN models. model
Source sentence
It was found that both the LSTM and the GINN models reached convergence after training for 300 epochs.
Data
- NASDAQ Composite (IXIC) market-prices public
Source sentence
The NASDAQ Composite (IXIC) dated from 06/01/1992 to 05/31/2022 is used as the dataset for the parametric study.
Costs
Not applicable
Finds 2
- The GINN model achieved the best out‑of‑sample prediction accuracy, obtaining 16 first‑place, 18 second‑place and 4 third‑place rankings across 21 time‑series and metric combinations. predictive-accuracy
Source sentence
Overall, the proposed GINN model and the GINN-0 model achieved high performance in out-of-sample prediction accuracy, beating other models in most test categories. Combined, they scored 16 first places, 18 second places, and 4 third places out of a total of 21 time series data and performance metric combinations.
- A weight λ of 0.01 was identified as the optimal hyper‑parameter for maximizing the Coefficient of Determination (R2) on the testing set. performance
Source sentence
Through testing, we determined that λ = 0.01 would yield the best performance.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 9 related papers). Coverage is partial, so this is neither replication nor refutation.
2 papers in the archive
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (44)
- Literature review: Machine learning techniques applied to financial market prediction (2019) — link
- Introduction to financial forecasting (1996) — link
- Forecasting the volatility of stock price index: A hybrid model integrating LSTM with multiple GARCH -type models (2018) — link
- Portfolio Selection (1952) — link
- A Hybrid Prediction Model Integrating GARCH Models With a Distribution Manipulation Strategy Based on LSTM Networks for Stock Market Volatility (2022) — link
- Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation (1982) — link
- Forecasting Volatility of Stock Indices with ARCH Model (2013) — link
- Generalized autoregressive conditional heteroskedasticity (1986) — link
and 36 more
- A forecast comparison of volatility models: does anything beat a GARCH (1,1)? (2005) — link
- Forecasting stock market volatility using (non-linear) Garch models (1996)
- Conditional Heteroskedasticity in Asset Returns : A New Approach (1991) — link
- On the Relation between the Expected Value and the Volatility of the Nominal Excess Return on Stocks (1993) — link
- Threshold heteroskedastic models (1994) — link
- A Continuous - Time GARCH Process Driven by a Lévy Process : Stationarity and Second - Order Behaviour (2004)
- The ZD - GARCH model: A new way to study heteroscedasticity (2018)
- Volatility Forecasting in the Hang Seng Index using the GARCH Approach (2009) — link
- Are the GARCH models best in out-of-sample performance? (1991) — link
- Forecasting UK stock market volatility (2000) — link
- A neural-network-based nonlinear metamodeling approach to financial time series forecasting (2009) — link
- An Empirical Comparison of Machine Learning Models for Time Series Forecasting (2010) — link
- A novel time series forecasting model with deep learning (2020) — link
- A comparison of artificial neural network and time series models for forecasting commodity prices (1996) — link
- An artificial neural network (p,d,q) model for timeseries forecasting (2010) — link
- Neural network forecasting for seasonal and trend time series (2005) — link
- Layered neural networks as universal approximators (1997)
- Recurrent Neural Networks Are Universal Approximators (2006)
- Deep Learning for Time - Series Analysis (2017) — link
- Long Short -term Memory (1997) — link
- Application of Neural Network for Forecasting of Exchange Rates and Forex Trading (2012) — link
- A LSTM -based method for stock returns prediction: A case study of China stock market (2015) — link
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations (2019) — link
- Physics-informed machine learning (2021) — link
- Accelerating Thermal Simulations in Additive Manufacturing by Training Physics - Informed Neural Networks With Randomly Synthesized Data (2023) — link
- bashtage/arch: Release 6.1.0 (2023) — link
- PyTorch : an imperative style, high-performance deep learning library (2019)
- Batch normalization: accelerating deep network training by reducing internal covariate shift (2015)
- Deep Learning using Rectified Linear Units ( ReLU ) (2018)
- Dropout: A Simple Way to Prevent Neural Networks from Overfitting (2014)
- Decoupled Weight Decay Regularization (2017)
- Yahoo Finance - Stock Market Live , Quotes , Business & Finance News (2023) — link
- Steel, R . G . D ., and J . H . Torrie : Principles and Procedures of Statistics . ( With special Reference to the Biological Sciences .) McGraw - Hill Book Company , New York , Toronto , London 1960, 481 S ., 15 Abb .; 81 s 6 d (1962) — link
- Mathematical Statistics : Basic Ideas and Selected Topics , Volumes I - II Package (2015) — link
- Advantages of the mean absolute error ( MAE ) over the root mean square error ( RMSE ) in assessing average model performance (2005)
- Persistence in Variance , Structural Change , and the GARCH Model (1990) — link
0 of 44 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.