Option Pricing under Accuracy–Cost Trade-Offs: A Multi-Criteria Framework Based on Monte Carlo Simulation, Machine Learning, and Data Envelopment Analysis

Document Type : Original Article

Author
Department of Mathematics, Faculty of Engineering Science, Quchan University of Technology, Quchan, Iran.
Abstract
Option pricing remains a fundamental challenge in financial engineering, particularly where pricing accuracy must be balanced against computational efficiency. While machine learning has shown promise in derivative valuation, selecting an appropriate pricing model remains difficult when multiple performance criteria must be considered simultaneously. This study proposes a multi-criteria model-selection framework for European option pricing that integrates Monte Carlo simulation, machine learning, and Data Envelopment Analysis (DEA). Using a geometric Brownian motion framework calibrated to realistic market conditions, we generate a comprehensive dataset of European call option prices under multiple volatility regimes and maturity structures. Four machine learning models—Artificial Neural Networks, Random Forest, Support Vector Regression, and Gradient Boosting Machines—are evaluated with respect to pricing accuracy, computational time, predictive stability, and overall efficiency. DEA is employed to construct a Pricing Efficiency Index that jointly evaluates prediction quality and computational cost. Results indicate that while Gradient Boosting achieves the highest predictive accuracy in several scenarios, Random Forest consistently provides the most attractive balance between pricing precision and computational burden. Robustness analyses across alternative volatility regimes and simulation settings confirm the stability of these findings. The results suggest that model selection in option pricing should not rely solely on predictive accuracy but should also account for computational efficiency, particularly in algorithmic trading, real-time risk management, and large-scale portfolio valuation. The framework offers a practical decision-support tool for financial engineers and quantitative analysts seeking efficient pricing methodologies under realistic computational constraints.
Keywords


Articles in Press, Accepted Manuscript
Available Online from 22 November 2026