Paper: arXiv 2505.19617
Authors: Dominik Stempień, Robert Ślepaczuk
Abstract
This research systematically develops and evaluates various hybrid modeling approaches by combining traditional econometric models (ARIMA and ARFIMA models) with machine learning and deep learning techniques (SVM, XGBoost, and LSTM models) to forecast financial time series. The empirical analysis is based on two distinct financial assets: the S&P 500 index and Bitcoin. By incorporating over two decades of daily data for the S&P 500 and almost ten years of Bitcoin data, the study provides a comprehensive evaluation of forecasting methodologies across different market conditions and periods of financial distress. Models’ training and hyperparameter tuning procedure is performed using a novel three-fold dynamic cross-validation method. The applicability of applied models is evaluated using both forecast error metrics and trading performance indicators. The obtained findings indicate that the proper construction process of hybrid models plays a crucial role in developing profitable trading strategies, outperforming their individual components and the benchmark Buy&Hold strategy. The most effective hybrid model architecture was achieved by combining the econometric ARIMA model with either SVM or LSTM, under the assumption of a non-additive relationship between the linear and nonlinear components.
Complexity vs Empirical Score
- Math Complexity: 7.0/10
- Empirical Rigor: 9.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical concepts like fractional integration (ARFIMA) and combines them with deep learning (LSTM) and hybridization techniques, indicating high mathematical complexity. It demonstrates high empirical rigor by using over two decades of daily data for S&P 500 and nearly ten years for Bitcoin, implementing a novel three-fold dynamic cross-validation, and evaluating both forecast accuracy and real-world trading performance with multiple risk-adjusted metrics.
Research Flowchart
flowchart TD A["Research Goal: Develop & Evaluate Hybrid Models for Financial Forecasting"] B["Data: S&P 500 (20+ yrs) & Bitcoin (10+ yrs)"] C["Methodology: 3-Fold Dynamic Cross-Validation & Hyperparameter Tuning"] D["Models: Econometric + ML/DL<br>(ARIMA/ARFIMA + SVM/XGBoost/LSTM)"] E["Non-Additive Hybrid Construction"] F["Evaluation: Forecast Errors & Trading Metrics"] G["Key Findings: Hybrid Models Outperform Benchmarks<br>Best: ARIMA + (SVM or LSTM)"] A --> B B --> C C --> D D --> E E --> F F --> G
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 hybrid model architectures that combine the econometric ARIMA model with either SVM or LSTM, assuming a non‑additive relationship between linear and nonlinear components, to improve forecasting and trading performance.
Source sentence
The most effective hybrid model architecture was achieved by combining the econometric ARIMA model with either SVM or LSTM, under the assumption of a non-additive relationship between the linear and nonlinear components.
Assumes 6
- Econometric models assume linear, stationary, and normal distribution properties in the data. statistical
Source sentence
This group of methods usually adopts assumptions concerning the linear, stationary, and normal distribution properties present in the data (Shah et al. 2019).
- Hybrid models assume a non‑additive relationship between linear and nonlinear components. model
Source sentence
under the assumption of a non-additive relationship between the linear and nonlinear components.
- ARIMA requires the underlying series to be stationary or first transformed into a stationary form. statistical
Source sentence
In order to apply ARIMA methodology, the underlying series is required to be either stationary or to be first transformed into the stationary form (Kobiela et al., 2022).
- Trading signals are generated only when predicted returns exceed transaction costs. execution
Source sentence
To generate a buy signal, the predicted value of the next day’s return must exceed the transaction costs.
Show 2 more
- Closing price series is transformed into logarithmic returns for model input. data
Source sentence
To generate input for the predictive models, the closing price series was transformed into the logarithmic returns.
- Transaction costs are modeled as a percentage parameter c in the signal rule. execution
Source sentence
c – transaction costs (formulated in percentages)
Data
2018-01 – 2023-12 · Bitcoin · crypto · daily
- S&P 500 daily price data market-prices unknown
Source sentence
... The data period for S&P 500 spanned between 1 January 2002 and 31 December 2023.
- Bitcoin daily price data crypto-onchain unknown
Source sentence
For Bitcoin, the data covered daily observations between 1 January 2015 and 31 December 2023.
Yahoo Finance
Costs
Modeled
c is a percentage representing transaction costs; costs are applied when generating trading signals and double‑charged when switching positions.
Source sentence
... The process of opening a new position or modifying an existing one is performed only if potential profits from the trade surpass the transaction costs.
Finds 5
annual return 30.0% · max drawdown 0.6%
- Hybrid ARIMA‑SVM and ARIMA‑LSTM models outperform their individual components and the Buy&Hold benchmark. performance
Source sentence
The obtained findings indicate that the proper construction process of hybrid models plays a crucial role in developing profitable trading strategies, outperforming their individual components and the benchmark Buy&Hold strategy.
- For Bitcoin, the LSTM‑ARIMA (1) hybrid achieves IR = 0.50 and SR = 1.06, surpassing both the market and its constituent models. performance
Source sentence
Long‑Short trading signals applied to the predictions of LSTM‑ARIMA (1) resulted in IR and SR metrics at the levels of 0.50 and 1.06, respectively.
- Hybrid SVM‑ARIMA (1), SVM‑ARIMA (2) and LSTM‑ARFIMA (1) also outperform the market and their individual models for Bitcoin. performance
Source sentence
The SVM‑ARIMA (1), SVM‑ARIMA (2), and LSTM‑ARFIMA (1) also managed to outperform both the market and their individual underlying models.
- XGBoost‑based hybrid approaches do not improve predictive power and underperform relative to other hybrids. negative
Source sentence
These observations are consistent with the findings for the S&P 500 index: the unsuitability of XGBoost‑based approaches and the superior performance of hybrid methodology not assuming an additive relationship between linear and nonlinear components.
- For Bitcoin, the ARIMA model yields the most accurate predictions according to RMSE and MAE, while hybridization does not improve accuracy. predictive-accuracy
Source sentence
The most accurate predictions, according to the RMSE and MAE metrics, were generated by the ARIMA model.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 37 related papers). Coverage is partial, so this is neither replication nor refutation.
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 (98)
- Predicting Prices Of S&P 500 Index Using Classical Methods and Recurrent Neural Networks (2024) — link
- Data-driven price trends prediction of Ethereum: A hybrid machine learning and signal processing approach (2024) — link
- LSTM-ARIMA as a hybrid approach in algorithmic investment strategies — → 2406.18206
- Machine learning Ethereum cryptocurrency prediction and knowledge-based investment strategies (2024) — link
- RNN-LSTM: From applications to modeling techniques and beyond—Systematic review (2024) — link
- Supervised autoencoder MLP for financial time series forecasting — → 2404.01866
- Return prediction by machine learning for the Korean stock market (2023) — link
- Forecasting cryptocurrencies volatility using statistical and machine learning methods: A comparative study (2023) — link
and 90 more
- Application of machine learning in algorithmic investment strategies on global stock markets (2023) — link
- A review on drawdown risk measures and their implications for risk management (2022) — link
- LSTM in Algorithmic Investment Strategies on BTC and S&P500 Index (2022) — link
- Applying Hybrid ARIMA-SGARCH in Algorithmic Investment Strategies on S&P500 Index (2022) — link
- Stock Market Prediction Using Machine Learning Techniques: A Decade Survey on Methodologies, Recent Developments, and Future Directions (2021) — link
- Long memory and efficiency of Bitcoin during COVID-19 (2021) — link
- Efficiency in cryptocurrency markets: new evidence (2021) — link
- Forecasting gold price with the XGBoost algorithm and SHAP interaction values (2021) — link
- Forecasting mid-price movement of Bitcoin futures using machine learning (2021) — link
- Machine Learning and Traditional Econometric Models: A Systematic Mapping Study (2021) — link
- Predictions of bitcoin prices through machine learning based frameworks (2021) — link
- Portfolio optimization with return prediction using deep learning and machine learning (2021) — link
- Nvidia’s Stock Returns Prediction Using Machine Learning Techniques for Time Series Forecasting Problem (2021) — link
- Stock market movement forecast: A Systematic review (2020) — link
- Deep Learning for Price Movement Prediction Using Convolutional Neural Network and Long Short-Term Memory (2020) — link
- Fractional Neuro-Sequential ARFIMA-LSTM for Financial Market Forecasting (2020) — link
- Adaptive market hypothesis: An empirical analysis of time –varying market efficiency of cryptocurrencies (2020) — link
- Beyond Econometrics: A Roadmap Towards Financial Machine Learning (2019) — link
- Volatility in the Cryptocurrency Market (2019) — link
- Combining Principal Component Analysis, Discrete Wavelet Transform and XGBoost to trade in the financial markets (2019) — link
- Time Series Classification of Cryptocurrency Price Trend Based on a Recurrent LSTM Neural Network (2019)
- Momentum and contrarian effects on the cryptocurrency market (2019) — link
- Stock Market Analysis: A Review and Taxonomy of Prediction Techniques (2019) — link
- An Empirical Study of Machine Learning Algorithms for Stock Daily Trading Strategy (2019) — link
- Public Mood–Driven Asset Allocation: the Importance of Financial Sentiment in Portfolio Management (2018) — link
- A New Hybrid Method for China’s Energy Supply Security Forecasting Based on ARIMA and XGBoost (2018) — link
- Predicting the Price of Bitcoin Using Machine Learning (2018) — link
- Advances in Financial Machine Learning: Lecture 10/10 (Presentation Slides) (2018) — link
- Machine Learning Methods in Algorithmic Trading Strategy Optimization – Design and Time Efficiency (2018) — link
- Deep learning with long short-term memory networks for financial market predictions (2017) — link
- Time-varying long-term memory in Bitcoin market (2017) — link
- Stock market one-day ahead movement prediction using disparate data sources (2017) — link
- A deep learning framework for financial time series using stacked autoencoders and long-short term memory (2017) — link
- Bridging the divide in financial market forecasting: machine learners vs. financial economists (2016) — link
- XGBoost: A Scalable Tree Boosting System (2016) — link
- Backtest Overfitting in Financial Markets (2016)
- The Future of Empirical Finance (2015) — link
- Recurrent neural network and a hybrid model for prediction of stock returns (2015) — link
- LSTM: A Search Space Odyssey (2015) — link
- Forecasting stock index returns using ARIMA-SVM, ARIMA-ANN, and ARIMA-random forest hybrid models (2014) — link
- A hybrid ARFIMA and neural network model for electricity price prediction (2014) — link
- Stock index forecasting based on a hybrid model (2012) — link
- Improvement in Forecasting Accuracy Using the Hybrid Model of ARFIMA and Feed Forward Neural Network (2012) — link
- Examining Sharp, Sortino and Sterling Ratios in Portfolio Management, Evidence from Tehran Stock Exchange (2011) — link
- Forecasting stock markets using wavelet transforms and recurrent neural networks: An integrated system based on artificial bee colony algorithm (2011) — link
- Calculating and Comparing Security Returns is Harder than you Think: A Comparison between Logarithmic and Simple Returns (2010) — link
- Modelling Exchange Rate Returns Using Non-linear Models (2010) — link
- Surveying stock market forecasting techniques - Part II: Soft computing methods (2009) — link
- Short term forecasting with support vector machines and application to stock price prediction (2008) — link
- Long memory in the Portuguese stock market (2007) — link
- Dependence and mean reversion in stock prices: The case of the MENA region (2006) — link
- An empirical investigation of the usefulness of ARFIMA models for predicting macroeconomic and financial time series (2006) — link
- A hybrid ARIMA and support vector machines model in stock price forecasting (2005) — link
- Rational bubbles or persistent deviations from market fundamentals (2005) — link
- Forecasting stock market movement direction with support vector machine (2005) — link
- 2005 Special Issue: A comparative study of autoregressive neural network hybrids (2005) — link
- Model-order selection: a review of information criterion rules (2004) — link
- The Efficient Market Hypothesis and Its Critics (2003) — link
- Financial time series forecasting using support vector machines (2003) — link
- Learning Precise Timing with LSTM Recurrent Networks (2003) — link
- Financial Forecasting Using Support Vector Machines (2001) — link
- Support vector machines: hype or hallelujah? (2000) — link
- Learning to Forget: Continual Prediction with LSTM (2000) — link
- Long memory in the Greek stock market (2000) — link
- Combined forecasts from linear and nonlinear time series models (1999) — link
- A Brief Introduction to Boosting (1999)
- Long Short-Term Memory (1997) — link
- A decision-theoretic generalization of on-line learning and an application to boosting (1997) — link
- Varieties of long memory models (1996) — link
- Support-Vector Networks (1995) — link
- A search for long memory in international stock market returns (1995) — link
- Long-Term Memory in Stock Market Prices (1991) — link
- Forecasting, Structural Time Series Models and the Kalman Filter (1990) — link
- The Principles of the Box-Jenkins Approach (1975) — link
- Spurious regressions in econometrics (1974) — link
- Time series analysis, forecasting and control (1972) — link
- Understanding support vector machine regression (2025)
- Cryptocurrency price forecasting–a comparative analysis of ensemble learning and deep learning methods (2024)
- ARIMA vs LSTM on NASDAQ stock exchange data (2022) — link
- Prediction-Based Portfolio Optimization Models Using Deep Neural Networks (2020) — link
- Fore-casting stock market crisis events using deep and statistical machine learning techniques (2018)
- STOCK-MARKET FORECASTING USING MACHINE LEARNING (2017)
- PSEUDO-MATHEMATICS AND FINANCIALCHARLATANISM: THE EFFECTS OF BACKTEST OVERFITTING ONOUT-OF-SAMPLE PERFORMANCE (2014) — link
- MUTUAL FUND PERFORMANCE* (2007) — link
- 25 years of time series forecasting (2006) — link
- To combine or not to combine: selecting among forecasts and their combinations (2005) — link
- Time series forecasting using a hybrid ARIMA and neural network model (2003) — link
- AN INTRODUCTION TO LONG‐MEMORY TIME SERIES MODELS AND FRACTIONAL DIFFERENCING (1980) — link
- Efficient capital markets (1970)
- International Journal of Uncertainty, Fuzziness and Knowledge-based Systems F the Vanishing Gradient Problem during Learning Recurrent Neural Nets and Problem Solutions
2 of 98 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.