Julius Blum, GmbH – Hierarchical Forecasting with Facebook’s Prophet Model

June 14th, 10.15-10.45 am

Our use-case provides a comparison of a classical Bottom-Up approach to forecasting grouped time series vs. an Ordinary Least Squares (OLS) reconciliation approach using Facebook’s state-of-the-art Prophet model. The study was conducted using more than 11000 grouped time series with a monthly frequency spanning ten years. Multiple measures of error computed over a test set were used to compare the results of both methods.



Dr. Josh Beal, Data Scientist | Julius Blum GmbH

picture of Dr. Josh Beal

Originally from USA. Background is Mathematics and Statistics. Received PhD in Mathematics in 2013 and was a Prof at University in Indiana until moving to Austria to work as Data Scientist for Julius Blum, GmbH.