Comparison of Feature Combinations on Simultaneous Prediction of Stock Price and Volatility

Wenhan Lu

2025

Abstract

Accurate prediction of stock prices and volatility is crucial for informed financial decision-making. However, traditional models often focus on single-target forecasts, neglecting the connection between price movements and volatility, which can limit predictive accuracy. Therefore, there is a need for more effective approaches that can simultaneously predict both stock prices and volatility. This study proposes an innovative method to address these challenges by using two target variables: the 5-day-ahead closing price and the 5-day high-low price difference as a measure of volatility. Besides, the study applies three feature selection techniques—Random Forest, Lasso Regression, and Mutual Information to identify the best features for predicting closing prices, which are then used to forecast volatility. The results of this study, based on data from Amazon, Google, and Microsoft over a 10-year period (2015-2025), show that Lasso Regression outperforms the other methods. It achieved the lowest mean squared error (MSE) across all three companies (Amazon: 0.2485; Google: 0.0323; Microsoft: 5.1805) while maintaining high R² values (above 0.78). The findings highlight Lasso Regression's ability to balance prediction accuracy and generalizability, offering a computationally efficient method for multi-target prediction, which improves the practicality of multi-target models for financial applications.

Paper Citation

in Harvard Style

Lu W. (2025). Comparison of Feature Combinations on Simultaneous Prediction of Stock Price and Volatility. In Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE; ISBN 978-989-758-765-8, SciTePress, pages 430-435. DOI: 10.5220/0013698900004670

in Bibtex Style

@conference{icdse25,
author={Wenhan Lu},
title={Comparison of Feature Combinations on Simultaneous Prediction of Stock Price and Volatility},
booktitle={Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE},
year={2025},
pages={430-435},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013698900004670},
isbn={978-989-758-765-8},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Science and Engineering - Volume 1: ICDSE
TI - Comparison of Feature Combinations on Simultaneous Prediction of Stock Price and Volatility
SN - 978-989-758-765-8
AU - Lu W.
PY - 2025
SP - 430
EP - 435
DO - 10.5220/0013698900004670
PB - SciTePress