Machine Learning for Quantitative Investing

On a periodic basis, publicly traded companies report fundamentals, financial data including revenue, earnings, debt, among others. When evaluating companies as potential investments, we can look at their pasts. We can compare them to other companies and consider how they performed as investments. But, ultimately, how companies evolve and how their share prices perform will be driven by future developments that we cannot observe in advance. In this talk, we discuss lookahead factor models that use deep neural nets to forecast future fundamentals. We use this forecast to build factor models instead of using historical fundamentals. Additionally, we estimate the uncertainty in our forecast and use them to reduce risk. We show, through simulation, that using forecast fundamentals and estimated uncertainty far outperform traditional factor models in systematic investing. Speaker: Lakshay Chauhan Lakshay is a Sr. ML Engineer at Euclidean, an investment management firm that focuses on leveraging machine learning for quantitative investing. Lakshay built the machine learning infrastructure and long-term forecasting models that power the investment models in production. Currently, he is focussing on using large language models for unstructured data in finance such as filings, documents, etc to improve investment models.