Feature selection methods on deep learning algorithms for stock price forecasting: A systematic literature review
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, cilt.181, sa.3, ss.1-30, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 181 Sayı: 3
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.engappai.2026.115471
- Dergi Adı: ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
- Derginin Tarandığı İndeksler: Applied Science & Technology Source, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Scopus, Science Citation Index Expanded (SCI-EXPANDED), Compendex, INSPEC
- Sayfa Sayıları: ss.1-30
- Erciyes Üniversitesi Adresli: Evet
Özet
Stock price forecasting is a difficult endeavor because of the extremely dynamic and nonlinear nature of the data. Deep learning methods have gained widespread usage as powerful tools for capturing complex temporal patterns and dependencies in financial time series, since deep architectures can autonomously extract useful features from raw, high-dimensional data. While feature selection is a well-established method for improving classical machine learning models, its role in deep learning remains ambiguous. This systematic review explores whether feature selection improves the predictive accuracy and efficiency of deep learning models in stock market forecasting. To clarify this, we carried out a thorough literature search across several scientific databases. We analyzed empirical outcomes from peer-reviewed papers published between 2010 and 2025, with a focus on strategies that combine feature selection techniques with deep learning models. This review relies on the main categories of feature selection methods as filters, wrappers, embeddings, and hybrid approaches while investigating their performances over various deep learning architectures, including long short-term memory networks, gated recurrent units, recurrent neural networks, convolutional neural networks, and transformer models. By analyzing trends from empirical reports, this study offers insights into the integration of feature selection methods with deep learning models to improve the performance of financial forecasting applications. In the dynamic field of quantitative finance, we also pinpoint research gaps and potential avenues for feature selection techniques on deep learning frameworks.