Застосування нейронних диференціальних рівнянь для прогнозування цін американських опціонів

Abstract

The paper considers the problem of forecasting American option prices using neural differential equations. Classical approaches to option pricing based on financial differential equations are analyzed, and their limitations under complex market dynamics are demonstrated. The use of neural differential equations is proposed as a method that combines the mathematical explanation of traditional models with the flexibility of machine learning techniques. The results show that training models on real and synthetic financial data can improve forecasting accuracy.