Quantitative Research Projects

A collection of quantitative finance and risk modeling projects demonstrating skills in time series forecasting, financial derivatives pricing, and credit risk assessment.

Projects Overview

1. Commodity Price Forecasting

Objective: Extrapolate monthly natural gas price data to daily granularity and forecast future prices for storage contract valuation.

Techniques:

• Time series decomposition (trend + seasonality)
• Sine/cosine cyclical feature engineering
• Linear regression for trend modeling
• Price interpolation and extrapolation

Key Files:

• commodity-price-forecasting/commodity-price-forecasting.ipynb
• commodity-price-forecasting/Nat_Gas.csv.csv

---

2. Storage Contract Valuation

Objective: Build a pricing model for natural gas storage contracts accounting for injection/withdrawal costs, storage fees, and seasonal price differentials.

Techniques:

• Cash flow analysis
• Contract valuation modeling
• Inventory constraint validation
• Multi-period optimization

Key Files:

• storage-contract-valuation/contract_pricing_model.ipynb

Results: Created a production-ready pricing function that calculates expected contract value given injection/withdrawal dates, volumes, and cost parameters.

---

3. Credit Risk Modeling

Objective: Predict probability of default for personal loan borrowers and calculate expected losses.

Techniques:

• Logistic Regression
• Random Forest Classifier
• Model comparison (Accuracy, ROC-AUC, Log Loss)
• Expected loss calculation (PD × EAD × LGD)

Key Files:

• credit-risk-modeling/credit-risk-modeling.ipynb
• credit-risk-modeling/Task 3 and 4_Loan_Data.csv

Results: Achieved 99.9% accuracy with ROC-AUC of 0.9999 in predicting loan defaults. Built function to estimate expected loss for any borrower profile.

---

4. FICO Score Bucketing

Objective: Develop optimal bucketing strategy to convert continuous FICO scores into categorical risk ratings.

Techniques:

• Quantization optimization
• Mean Squared Error (MSE) minimization
• Log-Likelihood maximization
• Dynamic programming approach

Key Files:

• bucket_generator/bucket_generator.ipynb
• bucket_generator/Task 3 and 4_Loan_Data.csv.csv

Results: Implemented two bucketing methods with comparative analysis, creating optimal risk rating system for mortgage portfolio.

---

Technical Stack

Languages: Python 3.11+

Core Libraries:

• Data Analysis: pandas, numpy
• Machine Learning: scikit-learn
• Visualization: matplotlib
• Optimization: scipy