JPMC Quantitative Research Job Simulation Programme

Conducted via Forage between July 2024 to August 2024.

Overview

This repository contains my work for the JP Morgan Chase & Co. Quantitative Research Job Simulation hosted on Forage. The simulation consists of 4 tasks, covering key areas in quantitative finance such as market forecasting, pricing models, credit risk analysis, and FICO score analysis. Each task is implemented in a dedicated notebook, and the corresponding datasets are provided.

Key Features

• Advanced Implementations: Beyond the task requirements, I implemented financial models such as Holt-Winters (for time series forecasting) and Logistic Regression (for credit risk analysis).
• Documentation: Task instructions, methodologies, and insights are documented within the notebooks and this README.
• Datasets: Includes Nat_Gas.csv, Task3n4_Loan_Data.csv, and Nat_Gas_forecast.csv for Tasks 1, 3, and 4 respectively.

---

Tasks

Task 1: Market Forecasting - Forecasting Natural Gas Market Price

Objective: Analyze historical natural gas price data (Nat_Gas.csv) to estimate past prices and forecast prices one year into the future.
Approach:

• Visualized the data to identify trends and seasonal patterns.
• Implemented Holt-Winters Exponential Smoothing for time series forecasting.
• Developed a function to take a date as input and return a price estimate.

Key Insights:

• Seasonal trends significantly impact natural gas prices.
• The model provides a reliable extrapolation for long-term storage contract pricing.

---

Task 2: Pricing Models - Pricing Natural Gas Contracts

Objective: Develop a prototype pricing model for natural gas contracts using forecasted data (Nat_Gas_forecast.csv).
Approach:

• Created a function to price contracts based on injection/withdrawal dates, rates, storage costs, and maximum volume.
• Assumed zero interest rates and no transport delays for simplicity.
• Tested the model with sample inputs to validate its accuracy.

Key Insights:

• The model generalizes well for various contract scenarios.
• Manual oversight is recommended before full automation.

---

Task 3: Credit Risk Analysis - Calculating Probability of Default (PD)

Objective: Build a model to predict the probability of default (PD) for loan borrowers using provided data (Task3n4_Loan_Data.csv).
Approach:

• Trained a Logistic Regression model to estimate PD.
• Incorporated a 10% recovery rate to calculate expected loss.
• Explored additional techniques (e.g., decision trees) for comparative analysis.

Key Insights:

• Logistic Regression provided a robust baseline for PD estimation.
• The model can be extended with more advanced techniques for improved accuracy.

---

Task 4: FICO Score Analysis - Bucketing FICO Scores

Objective: Create a rating map to bucket FICO scores into categories, where lower ratings signify better credit scores.
Approach:

• Applied quantization techniques to optimize bucket boundaries.
• Explored optimization criteria such as mean squared error and log-likelihood.
• Developed a generalizable approach for future datasets.

Key Insights:

• The bucketing strategy effectively summarizes FICO scores for model input.
• The approach can be adapted for different datasets and bucket sizes.

---

Notes

• The notebooks are provided for better visualization and understanding of the work. Python code prototypes are not included in this repository.
• Task instructions and methodologies are documented within the notebooks and this README.

---

Thank you for visiting! Feel free to reach out with any questions or feedback.