Online Program

Return to main conference page
Saturday, October 20
Sat, Oct 20, 7:30 AM - 8:30 AM
Hall of Mirrors
Continental Breakfast and Speed Poster 4 sponsored by Statgraphics

Developing an Open-Source, Self-Service Time Series Forecasting API with R and Microsoft Azure (304994)

*Alexandria Giese, Johnson & Johnson 
Julia Schoenewald, Johnson & Johnson 

Keywords: Timeseries, Forecasting, API, R package, Rshiny

At Johnson & Johnson, the need for timeseries forecasting spanned multiple business units including forecasting consumer product sales and server utilization for mobile applications. In order to meet this diverse set of needs and consistently produce quality forecasts, our team decided to build a self-service API. The API structure allowed us to develop one robust forecasting system that can be applied across business units.

Our team encountered several challenges when developing this API from a design, statistical, and code perspective. To overcome these challenges, we developed a custom R package containing the code for both timeseries forecasting and management of the API service. When paired with Microsoft Azure, the structure of an R package streamlined testing, documentation, and deployment of the system. Additionally, the team developed an RShiny UI to provide non-technical users an interface with the API.

Our team’s development of an open-source API, R package, and Shiny application customized to Johnson & Johnson’s needs taught us to best leverage these technologies and avoid cost and transparency issues previously encountered with commercial software solutions.