Paper: arXiv 2501.17366
Authors: Prashant Pilla, Raji Mekonen
Abstract
With the volatile and complex nature of financial data influenced by external factors, forecasting the stock market is challenging. Traditional models such as ARIMA and GARCH perform well with linear data but struggle with non-linear dependencies. Machine learning and deep learning models, particularly Long Short-Term Memory (LSTM) networks, address these challenges by capturing intricate patterns and long-term dependencies. This report compares ARIMA and LSTM models in predicting the S&P 500 index, a major financial benchmark. Using historical price data and technical indicators, we evaluated these models using Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The ARIMA model showed reasonable performance with an MAE of 462.1, RMSE of 614, and 89.8 percent accuracy, effectively capturing short-term trends but limited by its linear assumptions. The LSTM model, leveraging sequential processing capabilities, outperformed ARIMA with an MAE of 369.32, RMSE of 412.84, and 92.46 percent accuracy, capturing both short- and long-term dependencies. Notably, the LSTM model without additional features performed best, achieving an MAE of 175.9, RMSE of 207.34, and 96.41 percent accuracy, showcasing its ability to handle market data efficiently. Accurately predicting stock movements is crucial for investment strategies, risk assessments, and market stability. Our findings confirm the potential of deep learning models in handling volatile financial data compared to traditional ones. The results highlight the effectiveness of LSTM and suggest avenues for further improvements. This study provides insights into financial forecasting, offering a comparative analysis of ARIMA and LSTM while outlining their strengths and limitations.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 8.0/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper uses standard machine learning concepts without deep mathematical derivations, fitting a low-complexity profile, but it provides specific backtest-ready metrics (MAE, RMSE) and comparative results on historical S&P 500 data, showing high empirical rigor.
Research Flowchart
flowchart TD
A["Research Goal: Forecast S&P 500<br>using LSTM vs. ARIMA"] --> B["Data & Inputs<br>Historical Prices & Technical Indicators"]
B --> C{"Model Selection & Processing"}
C --> D["ARIMA Model<br>Linear Processing"]
C --> E["LSTM Model<br>Non-linear Processing"]
D --> F["Evaluation: MAE, RMSE"]
E --> F
F --> G["Key Findings/Outcomes"]
G --> H["ARIMA: MAE 462.1<br>RMSE 614<br>Accuracy 89.8%"]
G --> I["LSTM: MAE 369.32<br>RMSE 412.84<br>Accuracy 92.46%"]
G --> J["LSTM (No Features):<br>MAE 175.9, RMSE 207.34<br>Accuracy 96.41%"]
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 comparative study of ARIMA and LSTM models for forecasting the S&P 500 index.
Source sentence
This report compares ARIMA and LSTM models in predicting the S&P 500 index, a major financial benchmark.
Assumes 4
- ARIMA assumes stationarity and linear relationships in the data. model
Source sentence
These limitations are the reason why it is less effective with financial data as it is non-linear, has some long-range dependencies, and is influenced by many factors.
- ARIMA model uses only the SPX closing price as a single time series. data
Source sentence
For the ARIMA model, only the SPX closing price was used as it only works with just one time series and doesn’t take other features into account.
- Data splitting is done chronologically to avoid look‑ahead bias. execution
Source sentence
For ARIMA: The dataset is split into 80% training to fit the ARIMA model capturing (p, d, q), and 20% testing to validate predictions.
- Features are selected based on correlation magnitude (|corr| > 0.5). data
Source sentence
Features that showed strong positive or negative correlations were prioritized, with absolute values greater than 0.5, ensuring the LSTM model used relevant predictive information.
Data
- SPX closing price and technical indicators market-prices unknown
Source sentence
Using historical price data and technical indicators, we evaluated these models...
Bloomberg
Costs
Not applicable
Finds 3
- ARIMA achieved MAE 462.1, RMSE 614, and 89.8% accuracy, capturing short‑term trends but limited by linear assumptions. performance
Source sentence
The ARIMA model showed a reasonable performance with the results showing an MAE value of 462.1 and RMSE value of 614. The results showed an accuracy percentage of 89.8%.
- LSTM with favorable features achieved MAE 369.32, RMSE 412.84, and 92.46% accuracy, capturing short‑ and long‑term dependencies. performance
Source sentence
For the model that included favorable features (Figure 3), it achieved an MAE of 369.32 and an RMSE of 412.84. This showed an accuracy of 92.46%.
- LSTM without additional features performed best with MAE 175.9, RMSE 207.34, and 96.41% accuracy. performance
Source sentence
The model that did not include features (Figure 4) showed an MAE of 175.9 with
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 41 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.