Project Description: This project focus on using machine learning algorithm on fundamental data to predict the bankruptcy of the company next year. The finding were then used in constructing a portfolio that was held for 7 year period, and tested for quantitative analysis to validate if the method is a valid trading strategy.
A large portion of the project focuses on feature engineering and data cleaning, working with Compustat dataset, the data was engineered to fit the xgboost, catboost, logistic regression predictive models.
The portfolio was constructed as a equal-weighted portfolio long-short portfolio and regressed against S&P500 index to see if the trading strategy is valid and tested if Alpha is significant.