4 CONCLUSIONS
The study evaluates the performance of 4 feature
combinations in the simultaneous prediction of stock
closing prices and volatility.A decade-long dataset of
stock market records from 3 companies (Amazon,
Google, and Microsoft) was analyzed. To address
redundancy issues inherent in multi-objective
forecasting frameworks, two distinct target variables
were established: 1) the closing price after 5 days, and
2) the difference between the highest and lowest
prices after 5 days. Given the relative importance of
closing price prediction compared to volatility
forecasting and to reduce parameter bias in multi-
target prediction models, conventional multi-output
approaches were abandoned in favor of a sequential
methodology. Instead, in the study, 3 feature selection
methods helped identify key features for closing price
prediction. Then these selected features were used as
inputs to predict price volatility.
Empirical results revealed company-specific
variations in optimal feature combinations for multi-
objective prediction. However, the feature
combination selected through Lasso Regression
consistently demonstrated superior predictive
performance across all companies compared to
alternative selection methods.
There are still some limitations of this paper. First,
the analytical scope was restricted to three established
feature selection techniques, potentially limiting
comprehensive exploration of the feature space.
Second, the Mutual Information and Random Forest
methods exhibited similar tendencies toward feature
selection, leading to repeated results in feature
combinations.
Future research can build upon this work in
several directions. First, more diverse feature
selection methods could be incorporated, particularly
those leveraging automatic feature extraction
techniques integrated with deep learning. Second,
alternative evaluation metrics, such as return-based
assessments, could be adopted to improve the
practical applicability and robustness of the model.
These avenues of research have the potential to
further enhance the precision of multi-target
prediction models, providing valuable support for
financial decision-making
REFERENCES
Blitsi, A. K. (2024). Multi-target learning with constraints
[Doctoral dissertation, Aristotle University of
Thessaloniki].
Htun, H. H., Biehl, M., & Petkov, N. (2023). Survey of
feature selection and extraction techniques for stock
market prediction. Financial Innovation, 9, 26.
Iranzad, R., & Liu, X. (2024). A review of random forest-
based feature selection methods for data science
education and applications. International Journal of
Data Science and Analytics, 1-15.
Jovi, A., Brki, K., & Bogunovi, N. (2015, May). A review
of feature selection methods with applications. In 2015
38th International Convention on Information and
Communication Technology, Electronics and
Microelectronics (MIPRO) (pp. 1200-1205). IEEE.
Kursa, M. B., & Rudnicki, W. R. (2011). The all relevant
feature selection using random forest. arXiv preprint
arXiv:1106.5112.
Muthukrishnan, R., & Rohini, R. (2016, October). LASSO:
A feature selection technique in predictive modeling for
machine learning. In 2016 IEEE International
Conference on Advances in Computer Applications
(ICACA) (pp. 18-20). IEEE.
Nabipour, M., Nayyeri, P., Jabani, H., Shahab, S., &
Mosavi, A. (2020). Predicting stock market trends using
machine learning and deep learning algorithms via
continuous and binary data; a comparative analysis.
IEEE Access, 8, 150199-150212.
Nguyen, X. V., Chan, J., Romano, S., & Bailey, J. (2014,
August). Effective global approaches for mutual