Analyzing the critical steps in deep learning-based stock forecasting: a literature review

Area of Science:

• Financial markets and computational intelligence.
• Machine learning applications in econometrics.

Background:

• Stock market forecasting faces challenges due to market uncertainty and dynamic conditions.
• Traditional analysis methods struggle with inherent financial market unpredictability.
• Deep learning models are increasingly explored for enhanced stock prediction accuracy.

Purpose of the Study:

• To systematically review deep learning-based stock forecasting models.
• To investigate the impact of critical implementation steps on forecasting performance.
• To identify research gaps and provide guidance for future studies in financial forecasting.

Main Methods:

• Systematic literature search across three databases for studies from 2020-2024.
• Analysis of influential studies focusing on seven critical steps for model development.
• Summarization of findings in tables and detailed identification of literature gaps.

Main Results:

• Deep learning models show promise but require careful execution of key steps.
• The review identified specific challenges and best practices in data handling, feature engineering, and model selection.
• Critical steps significantly influence the accuracy and reliability of stock forecasting models.

Conclusions:

• A systematic approach to data collection, feature engineering, and model selection is crucial for effective deep learning-based stock forecasting.
• This review highlights areas for improvement and offers a roadmap for developing more robust financial prediction models.
• Further research should focus on addressing identified literature gaps to advance the field of algorithmic trading and financial analytics.