This project uses quantitative research methods, machine learning, and data analysis to solve challenges in the financial and energy sectors. The analysis focuses on developing pricing models for natural gas contracts, estimating loan default probabilities, and forecasting losses for banks. Techniques such as K-means clustering, machine learning methods, and time series decomposition were employed to provide actionable insights.
Project highlights:
Development of a prototype pricing model for natural gas contracts, based on estimated prices for buying/selling on specific dates, gas injection/withdrawal rates in storage, and associated costs.
Estimation of loan default probability and expected loss for a bank by analyzing a loan portfolio using machine learning techniques.
Use of quantization to categorize FICO scores into reduced buckets, facilitating data handling and ensuring model applicability on future datasets.