Predictive Analytics for Covid-19 Recovery

Reopening the economy and reestablishing social contact are vital to the recovery of the economy and society from Covid-19\. Early reopening for economic stimulus risks having the opposite effect, as repeated restrictions imposed at short notice potentially do more cumulative damage, and for longer, than maintaining initial restrictions -- shops, offices and factories have invested in reopening and social distancing measures and individuals have made commitments on the expectation of being able to fulfil them.

It is inevitable that restrictions will need to come and go for the foreseeable future, given that the only readily available metrics for the impact of changes in policy are changes in the Covid-19 R-Factor and reported incidence and mortality. The need is for policy-makers in any sector is to be able to make informed and timely decisions that impact the least number of people in the smallest area for the shortest period of time.

In any sector -- government, healthcare, sports and leisure venues, retail malls, factories etc -- decisions are dependent on the quality and range of data available, on its geographical, demographic and sector detail and, crucially, on the ability to integrate multiple sources, identify relevant trends, anticipate what may happen next and make informed choices to continue, relax or reimpose restrictions.

This is where the problem arises: data is scattered across diverse sources, is of variable quality, accessibility, timeliness, completeness and accuracy, and curating it to generate effective local or sector insight is slow and labour-intensive, often using platforms that are themselves restrictive and/or expensive to operate.

This still only reflects what participants knew to look for -- it does not help surface previously unsuspected relationships that might then influence decision-making. Even then, such relationships need to be validated, but the biggest lag is in inspiration -- thinking to look for particular correlations. The pandemic has given us many examples: the correlation of ethnicity with mortality; the impact of living at altitude with severity or the propensity of Covid-19 patients to develop other conditions following apparent recovery.

There we are still simply identifying patterns, trends and relationships and driving specific metrics. All are useful, but prediction is usually by eyeball or simple projection of a trend.

Two Worlds is developing an SaaS service based on udu, a next-generation, AI-driven intelligence platform. The outcome is a system customisable to local or sector need and which dynamically integrates specific data sources with the automated discovery of supplementary data.

We bring a range of statistical, mathematical and AI approaches to the analysis and presentation of information and to the prediction of trends in data. Our approach enables both the creation of repeatable reporting and prediction and the self-organising discovery of new and potentially relevant patterns and relationships. In doing so, it builds on udu's established market, Two Worlds' prior (and continuing) environmental analytics and a first stage Covid-19 analytical study, now approaching completion.

This project enables the project to move from prototype demonstrator (TRL5) to being usable by initial test customers (TRLs 7+).