Spatiotemporal Forecasting of the Economic Effects of COVID19

The Problem:

Early in the COVID19 pandemic our client, an office in the Australian government, was concerned with the long term economic impacts of public health control measures. While some immediate effects were obvious, other long term effects were less clear. Given the nature of the data, considerations for both spatial and temporal dependencies needed to be taken into account. The client was seeking a configurable platform for a non-coding analyst to run their own forecasts using a combination of public data on COVID19 cases and economic indicators, and simulate alternative outcomes based on actionable scenarios. 

The Data:

Public data is available at the state and territory level for COVID19 cases, hospitalizations, and deaths. Additionally various public economic datasets were available at the same grain, including passenger transport, unemployment, manufacturing production, mining output, imports, exports, etc. We also incorporated Google's mobility data. 

The Solution:

We designed and deployed a web app that provided spatiotemporal forecasting and prescriptive simulation. We automated the ETL of the various public datasets, and configured the spatial referencing for Australian states and territories. The user had the ability to choose various input features, run a forecast, and interpret the results. They then were able to one or more features to manually modify, run a forecast given that scenario, and perform a gap analysis between the scenario and baseline forecasts.

The Impact:

Using our tool, the client was able to configure and run their own forecast and scenario models, and independently test various what if questions to guide policy decisions.