Model Development Strategies For Time Series Forecasting with American Express

Jun Kim - Director, Finance Decision Science, American Express

Abstract:

Jun will provide an overview of how his team has implemented multiple model development strategies to forecast time series data. The time series forecasting is different from other machine learning problems since every data point is attached to a time stamp. The model development strategies include directly using time series models (e.g., SARIMAX, Prophet) and also converting the time series dataset into a typical machine learning dataset to use traditional machine learning models (e.g., GBM). Jun will provide an example of how his team forecast AXP card members’ future spend on large merchants.

Bio:

Jun is a director on the Finance Decision Science team at American Express. He is in charge leading a team of data scientists to build predictive machine learning models to forecast key financial measures in order to add stability and predictability to the company’s P&L. Jun earned a B.S in Computer Science and Economics from University of Toronto and M.S. in Finance from University of Illinois at Urbana-Champaign. He is also a CFA charterholder and a host on the Value Investing Podcast.