Beyond Black-Scholes: A Synthesis of AI-Driven Advancements in Derivatives Pricing, Risk Sensitivity, and Computational Efficiency (2020–2025)
0 viewsAbstract
Traditional derivatives pricing models, while theoretically elegant, increasingly struggle with the scale, non-linearity, and real-time demands of modern financial markets. This paper provides a comprehensive synthesis of recent advancements in artificial intelligence (AI) and machine learning (ML) applied to the valuation of financial instruments. Through a systematic review of literature and empirical results from 2020 to 2025, the study evaluates diverse methodologies including Gaussian Process Regression (GPR), Physics-Informed Neural Networks (PINNs), and gradient boosting frameworks like CatBoost. The findings demonstrate that AI-based approaches achieve computational speed-ups of three to four orders of magnitude compared to traditional finite-difference and Monte Carlo methods, while maintaining relative error rates as low as 0.5% for complex instruments like variance swaps and American options. Furthermore, the paper examines the integration of "Greeks" estimation and deep calibration into unified AI frameworks, addressing the operational bottlenecks of risk management. While the "black box" nature of deep learning remains a primary concern for regulators and practitioners, the emergence of hybrid models—combining financial theory with machine learning acceleration—offers a path toward robust, interpretable, and near-instantaneous financial engineering. The study concludes that AI is not a replacement for traditional quantitative finance but a powerful extension that redefines what is analytically feasible in high-frequency environments.
Keywords
Artificial Intelligence (AI), Financial Risk Management, Machine Learning, Quantitative Finance