Criscent Birungi

About me

My doctoral research lies at the intersection of Mathematical Finance and Machine Learning. I recently used data from the U.S. Bureau of Labor Statistics to develop optimal annuitization strategies for individuals aged 60 and over. I am currently part of a research team at Desjardins Global Asset Management (DGAM) team designing a global stock recommendation system using AI models that integrate financial data and market trends. More broadly, my work spans mathematical and computational finance, statistical machine learning, econometrics, statistical theory, and time series analysis to develop machine learning tools to address challenges in finance, insurance, economics, and climate change.

Outputs

Recommending the right securities at the right time: A global system based on optimized decision-tree algorithms for dynamic factor solutions

This paper presents a dynamic stock recommendation system trained on a diverse dataset of fundamental and technical factors to extract the most important variables for return prediction. In collaboration with Desjardins Global Asset Management (DGAM), we employ efficient decision tree–based algorithms (XGBoost and LightGBM) with Bayesian optimization to handle computational constraints and diverse factor inputs. Based on these models, we implement and optimize a long/short trading strategy that takes long positions in the top-performing quintile and short positions in the bottom-performing quintile of ranked stocks. This study represents the first implementation of such machine learning frameworks within DGAM’s quantitative investment context. Our results demonstrate that this recommender system consistently delivers strong forecasting accuracy. While technical factors drive short-term rankings, fundamental factors ensure longer-term stability. The XGBoost-based strategies prove particularly robust, outperforming comparative models' cumulative returns and demonstrating return convexity that enhances performance across both favorable and adverse market environments. Despite a high monthly turnover, the strategy offers substantial diversification benefits and superior risk management. This study confirms that integrating decision-focused machine learning into portfolio construction enables investors to effectively align predictions with desired risk-return profiles.