Paper: arXiv 2309.00618

Authors: Albert Wong, Steven Whang, Emilio Sagre, Niha Sachin, Gustavo Dutra, Yew-Wei Lim, Gaetan Hains, Youry Khmelevsky, Frank Zhang

Abstract

Creating accurate predictions in the stock market has always been a significant challenge in finance. With the rise of machine learning as the next level in the forecasting area, this research paper compares four machine learning models and their accuracy in forecasting three well-known stocks traded in the NYSE in the short term from March 2020 to May 2022. We deploy, develop, and tune XGBoost, Random Forest, Multi-layer Perceptron, and Support Vector Regression models. We report the models that produce the highest accuracies from our evaluation metrics: RMSE, MAPE, MTT, and MPE. Using a training data set of 240 trading days, we find that XGBoost gives the highest accuracy despite running longer (up to 10 seconds). Results from this study may improve by further tuning the individual parameters or introducing more exogenous variables.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper primarily applies established machine learning models with standard metrics (RMSE, MAPE) and describes a backtesting setup with real market data and rolling windows, demonstrating high empirical rigor, but lacks advanced mathematical derivations or novel theoretical complexity.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Forecast Short-Term Stock Prices"] --> B["Data Collection<br>NYSE Stocks Mar 2020 - May 2022"]
  B --> C["Methodology<br>Train/Test Split 240 Days"]
  C --> D{"Model Training<br>& Tuning"}
  D --> E["XGBoost"]
  D --> F["Random Forest"]
  D --> G["MLP"]
  D --> H["SVR"]
  E & F & G & H --> I["Evaluation Metrics<br>RMSE, MAPE, MTT, MPE"]
  I --> J{"Outcomes"}
  J --> K["Winner: XGBoost<br>Highest Accuracy"]
  J --> L["Trade-off: Runtime<br>Up to 10 seconds"]
  K & L --> M["Conclusion<br>Potential for Further Tuning"]

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 to compare four machine learning models for short‑term stock price forecasting and reports that XGBoost achieves the highest accuracy.

Source sentence
we find that XGBoost gives the highest accuracy despite taking longer (up to 10 seconds) to run.

Assumes 2

  • A training data set of 240 trading days is sufficient for model evaluation. data
    Source sentence
    Using a training data set of 240 trading days, we find that XGBoost gives the highest accuracy despite taking longer (up to 10 seconds) to run.
  • Exogenous variables representing short‑term interest rate movement are relevant for forecasting. other
    Source sentence
    There are many possibilities of exogenous variables that can be used for a machine learning algorithm. For the purpose of this research, we have built on our previous study

Data

  • Three NYSE stocks market-prices unknown
    March 2020 to May 2022 · three well‑known NYSE stocks
    Source sentence
    forecasting three well‑known stocks traded in the NYSE in the short term over the period from March 2020 to May 2022

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • XGBoost produced the lowest forecasting errors on average compared to the other models. performance
    Source sentence
    the XGBoost models produced the lowest errors on average compared to the other machine learning models
  • Random Forest had the longest training time, followed by XGBoost. performance
    Source sentence
    it is noticeable from the MTT metric that the Random Forest algorithm takes the longest to train, followed by XGBoost
  • Increasing the number of estimators from 100 to 300 gave little to no improvement in MAPE for both XGBoost and Random Forest. performance
    Source sentence
    increasing the N‑estimators parameter from 100 to 300 for both the XGBoost and the Random Forest algorithm showed little to no signs of improvement with respect to the MAPE performance measure
  • All 5‑day‑ahead forecasts had higher error metrics than 1‑day‑ahead forecasts, as expected. predictive-accuracy
    Source sentence
    Overall, all values of performance metrics considered for the 5‑day ahead forecasts are higher than those for the 1‑day ahead forecasts
  • Models achieved MAPE of 5% or lower, demonstrating usefulness of the approach. predictive-accuracy
    Source sentence
    With MAPE at 5% or lower, these forecasting models demonstrated the usefulness of the approach taken in this research

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 30 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.