Price Forecasting & Credit Risk Models – JPMorgan Quantitative Research Simulation

Python-based predictive and financial modeling project built during the JPMorgan Chase & Co. Quantitative Research Virtual Experience (Forage).
Explore the code »

Report Bug · Request Feature

---

📌 About The Project

Price Forecasting & Credit Risk Models is a collection of Python-based tasks completed for the JPMorgan Chase & Co. Quantitative Research Virtual Experience Program (Forage, Nov 2025).
It showcases practical applications of time-series forecasting, quantitative pricing models, and credit risk prediction using machine learning.

Highlights

📈 Built linear regression models to forecast natural gas prices based on historical monthly data
📊 Designed a pricing engine for gas storage contracts, simulating injection, withdrawal, and cost parameters
🤖 Developed an XGBoost-based classifier to estimate loan default probability (PD) and compute expected loss with 90%+ accuracy
📉 Visualized feature importance and price trends to interpret model behavior

---

🔍 How It Works

Task 1 & 2 – Price Forecasting and Contract Valuation

• Loaded and preprocessed historical natural gas prices from CSV files
• Used Linear Regression to extrapolate monthly prices one year into the future
• Implemented calculate_contract() to simulate gas injection, withdrawal, and cost-based cash flows
• Visualized actual vs predicted price trends using Matplotlib

Task 3 – Credit Risk Modeling

• Processed borrower dataset and trained an XGBoostClassifier
• Evaluated model performance with Accuracy, F1 Score, and AUC metrics
• Created predict_loss_simple() to compute expected credit loss using PD and recovery rate assumptions
• Displayed feature importance and decision trees for interpretability

---

🚀 Built With

• 
• Pandas
• NumPy
• Scikit-learn
• XGBoost
• Matplotlib

🙏 Acknowledgments

• JPMorgan Chase & Co. – Quantitative Research Virtual Experience (Forage)

---

📷 Screenshots

XGBoost Model importance score - Analyzed a book of loans to estimate a customer's probability of default

Predicted Gas Prices using Linear Model

---