Graphical Models for Financial Time Series and Portfolio Optimization

Explore graphical models for constructing optimal portfolios in this 41-minute conference talk from the Toronto Machine Learning Series. Delve into various techniques including PCA-KMeans, autoencoders, dynamic clustering, and structural learning to capture time-varying patterns in covariance matrices. Compare the performance of portfolios generated using these graphical strategies against baseline methods and the S&P 500 index. Discover how these models consistently produced steadily increasing returns with low risk, often outperforming the market benchmark. Gain insights into the effectiveness of graphical models in learning temporal dependencies in time series data and their practical applications in asset management. Learn from Jenny Ni Zhan, a PhD student at Carnegie Mellon University, as she presents her research on leveraging machine learning techniques for financial portfolio optimization.