Projects
Standard portfolio optimization often suffers from estimation errors, leading to financial losses. While deep learning offers better pattern recognition, existing models like LSTMs are computationally heavy and complex, creating a need for faster, more stable alternatives.
We implemented Modern Hopfield Networks to capture complex market patterns efficiently. By replacing standard layers with Hopfield layers, this solution matches or beats state-of-the-art models like Transformers while being significantly faster and more stable to train.
Mastered advanced Deep Learning architectures: Hopfield Networks, Transformers and time-series embedding: Time2Vec. Gained expertise in quantitative finance metrics: Sharpe/Sortino ratios and rigorous backtesting strategies: Combinatorial Purged Cross-Validation.
This project offers a scalable tool for dynamic asset allocation and risk management. The architecture is flexible enough to incorporate unconventional data sources, such as ESG ratings or market sentiment, to enhance investment decision-making.
Corporate Credit Rating Forecast using Machine Learning Methods
Corporate credit ratings, issued by credit rating agencies like Standard & Poor's and Moody's, express the agency's opinion about the ability of a company to meet its debt obligations. Each agency applies its own methodology to measure creditworthiness and this assessment is an expensive and complicated process. Usually, the agencies take time to provide new ratings and update older ones. This causes delays in decision-making process for investors who use these ratings to assess their credit risk.
One solution to address delays would be to use the historical financial information of a company to build a predictive quantitative model capable of forecasting the credit rating that a company will receive. I employed machine learning techniques, creating classification models that quickly forecast credit ratings.
Explored classification methods like XGBoost, RandomForest and techniques to address imbalance in datasets - SMOTE. Also delved into financial ratios gaining knowledge on understanding a company's fiscal strength.
The insights gained can aid financial analysts, investors, and companies in making more informed and quick decisions related to credit risk. The classification methods used here can also be used to forecast other ratings like ESG Ratings.