WP-2026-002 · quant · ml · xai

Credit Risk & Options Pricing - A Comparative Study

Ibrahim Laklaa · 2026

classifiers

pricing methods

Abstract

Five classifiers (logistic regression, random forest, gradient boosting, XGBoost, neural networks) on synthetic credit data, side-by-side with three European call pricing methods: Black-Scholes analytical, Monte Carlo, and a neural-network approximator. Explainable-AI lens throughout.

Two problems, one lens

The study puts side by side two classic questions of quantitative finance, credit default prediction and European option pricing, and looks at both through explainability.

Credit risk

Five classifiers are trained on synthetic credit data: logistic regression, random forest, gradient boosting, XGBoost and a neural network. The logistic baseline reads

P (default ∣ x) = σ (β_{0} + β^{⊤} x), σ (z) = \frac{1}{1 + e ^{- z}}

Option pricing

A European call is priced three ways. The analytical Black-Scholes price:

C = S_{0} N (d_{1}) - K e^{- r T} N (d_{2}), d_{1, 2} = \frac{ln ( S _{0} / K ) + ( r \pm \frac{1}{2} σ ^{2} ) T}{σ T}

A Monte Carlo estimator under the risk-neutral measure:

C^= e^{- r T} \frac{1}{M} i = 1 \sum M (S_{T}^{(i)} - K)^{+}

And a neural network trained to approximate the pricing map.

Takeaway

The report compares accuracy and interpretability across all methods. Results and figures are in the PDF.

Read the paper Code