Artificial Intelligence in Financial Forecasting: Accuracy and Limitations

Abstract

This study critically investigates the evolving role of artificial intelligence (AI) in financial forecasting through a systematic literature review conducted across multiple reputable academic databases. The main objective is to assess the performance, interpretability, and practical integration of AI models within the financial domain. Using predefined inclusion and exclusion criteria, 43 peer-reviewed articles published between 2020 and 2025 were selected and thematically analyzed. Key AI techniques examined include machine learning, deep learning, and reinforcement learning, each demonstrating superior forecasting accuracy over traditional statistical methods. However, the study identifies persistent limitations, including model opacity, data quality concerns, and compliance challenges. A significant trade-off is observed between model accuracy and interpretability, particularly with complex deep learning models. Moreover, case studies highlight the practical success of AI in areas such as credit risk assessment, cash flow prediction, and portfolio optimization. The findings underscore the necessity for explainable AI (XAI) frameworks and human-AI collaboration to enhance trust and accountability in financial decision-making. The study concludes with recommendations for practitioners and policymakers to adopt transparent, auditable models and for researchers to focus on the development of robust, interpretable, and ethical AI-driven forecasting systems. This review contributes to the growing discourse on responsible AI adoption in finance and provides a foundation for future research and policy design.