Investigating the Effectiveness of Option Pricing Using Machine learning models Compared to the Black-Scholes model

One of the fundamental challenges investors face in capital markets is risk management. Options are considered one of the most practical financial instruments for risk management. Therefore, the pricing methods for these instruments hold particular significance. However, the complex and nonlinear relationships among the factors affecting option pricing have made modeling these relationships challenging, often leading to the use of restrictive and unrealistic assumptions in constructing models. One proposed solution to this issue is the application of machine learning algorithms. These algorithms have no limitations in identifying complex nonlinear relationships between variables and can build the required models without relying on unrealistic assumptions. In this context, the aim of this research is to evaluate the performance of machine learning algorithms in predicting option prices compared to the Black-Scholes model. This study utilized data from 144 options traded on the Tehran Stock Exchange between April 2018 and May 2024. The options were priced using both machine learning algorithms and the Black-Scholes model. To assess the performance of these models, their predictions were compared with the market prices of the options using statistical metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). Additionally, this research examined the performance of machine learning algorithms and the Black-Scholes model by considering the contract duration and the intrinsic value of the options. The results indicate that machine learning algorithms outperformed the Black-Scholes model in predicting option prices. Furthermore, the comparison of the models’ performance in predicting prices based on contract duration and intrinsic value also confirmed the superiority of machine learning algorithms. Among the models, the Gradient Boosting algorithm demonstrated the best performance compared to other methods.