Academic Project · La Trobe University
Built a Credit Risk Classification Model Using Machine Learning
Executive Summary
For my Master of Financial Analysis at La Trobe University, I built a credit risk classification model using logistic regression and random forest techniques to predict borrower risk. The project required cleaning and preparing financial and credit data, selecting and testing model features, and evaluating classification performance against real-world credit risk criteria. It sits alongside a related project on robo-advising and algorithmic trading, both applying predictive analytics directly to financial decision-making. Together, they demonstrate the data science depth behind the forecasting and dashboard work I deliver commercially.
Context
This was an applied academic project completed in 2025 during my Master of Financial Analysis at La Trobe University, focused on credit risk prediction using machine learning classification techniques.
Challenge
Credit risk classification requires balancing predictive accuracy against interpretability, particularly for financial decisions where a model's reasoning needs to be explainable to non-technical stakeholders. Feature selection and model comparison were central to building a model that performed reliably.
Strategic Approach
Phase 1
Data Preparation
Cleaned and prepared financial and credit data for classification modelling.
Phase 2
Model Development
Built and compared logistic regression and random forest classification models to predict credit risk.
Phase 3
Evaluation
Assessed model performance against real-world credit risk criteria to determine the most reliable classification approach.
Quantifiable Outcomes
- Built and compared two distinct classification approaches, logistic regression and random forest.
- Applied the same predictive analytics discipline to a companion robo-advising and algorithmic trading project.
- Extended core Python and predictive analytics skills directly into financial risk assessment.
Qualitative Achievements
- Demonstrated the data science depth behind the forecasting and dashboard work delivered commercially.
- Built a model designed to be explainable to non-technical financial stakeholders, not just accurate.
- Strengthened the Python and machine learning foundation now applied in commercial predictive analytics work.
A credit risk model is only as useful as its ability to be trusted by the person making the lending decision, which means interpretability matters as much as raw accuracy.
This project is the clearest evidence of the machine learning depth behind the commercial data work. It is the technical foundation, the commercial engagements are where that foundation gets applied.