Kathryn Hume — Financial Models, ML, and 17th-Century Philosophy

Forecasting Models

Kathryn Hume, Vice President of Digital Investments Technology at RBC, shares insights into building personal finance forecasting models. These models aim to predict upcoming payments, helping customers avoid overdraft fees by offering timely reminders. Kathryn explains, "We built something that's trying to predict upcoming payments in the next seven days," highlighting the model's practical application 1. The model's accuracy varies by payment type, achieving up to 98% accuracy for pre-approved payments, but facing challenges with more variable transactions like loans 2. The initiative reflects RBC's commitment to customer loyalty over profit from fees, aligning with their ethos of being a good corporate citizen 3.

   

Credit Assessment

In the realm of credit assessment, Kathryn discusses the application of machine learning to enhance decision-making processes. Traditional models, like linear regressions, are being upgraded with ML to better assess loan risks and optimize call center operations 4. Fairness in these models is a critical concern, with RBC implementing fairness tests before deploying models. Kathryn notes, "There's a lot of fairness tests prior to putting a model into production at the bank," emphasizing the importance of equitable treatment across different demographic groups 5.

   

Trade Execution

Kathryn explores the innovative use of reinforcement learning (RL) in optimizing trade execution. This approach helps distribute large stock orders throughout the day to minimize market impact, especially during volatile periods like March 2020 6. The RL model adapts to market changes, offering superior returns compared to traditional algorithms. Kathryn describes the process, "You might sell, you might buy, but, yeah, you're trading stocks during the day," illustrating the dynamic nature of RL in trading 7.