Reproduce “American Option Pricing using Self-Attention GRU and Shapley Value Interpretation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yanhui Shen
This research from the University of Wisconsin-Madison proposes and evaluates a self-attention GRU model for pricing SPY American-style call options, demonstrating its superior performance over traditional methods with RMSE improvements of over 50 times in some cases. The work uniquely integrates SHapley Additive exPlanations (SHAP) to interpret the model's predictions, revealing key drivers like current spot/strike prices and the significant influence of historical call prices.
Key insights
- Deep learning models, particularly the Self-Attention GRU, demonstrate substantial improvements in American option pricing accuracy and computational efficiency compared to traditional binomial tree methods.
- The inclusion of historical option prices, especially the previous day's call price, significantly enhances the predictive power of deep learning models for American options.
- SHAP analysis reveals that current spot price, strike price, and moneyness are the most impactful features, while also providing granular insights into the complex interplay and nuanced contributions of historical prices, volatility, and interest rates across different option types (OTM, ATM, ITM).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.