Systematic review of deep and machine learning for financial modeling

Machine learning (ML) and deep learning (DL) have impacted financial analytics, with advanced solutions for classification and regression tasks. This systematic review provides an analysis of state-of-the-art ML/DL applications in finance, with focus on methods, but also challenges. A total of 41 papers were analysed to identify trends, methodologies, and research gaps in this domain. The study begins with an overview of ML/DL in finance, main classification and regression problems, and challenges in financial data modelling. It then explores ML/DL techniques for classification tasks such as credit scoring, fraud detection, and algorithmic trading, evaluating traditional and modern approaches, including transformer-based models for sentiment analysis. Regression-oriented applications are analysed, with focus on stock price prediction, volatility forecasting, and portfolio optimization, with insights into hybrid modelling strategies. Comparative analysis assesses ML/DL models based on performance metrics, interpretability, and trade-offs between accuracy, computational complexity, and generalizability. The paper also identifies challenges, including data quality, ethical concerns, models, and the integration of ML/DL with traditional financial frameworks. Trends such as explainable AI, federated learning, and quantum computing are discussed as future directions. Findings show the increasing role of ML/DL in financial decision-making.