54368udu: AI Platform for Pandemic IntelligenceClosedFeasibility StudiesInnovate UKThe UK, amongst others, has been shown to lack a coherent and reliable infrastructure to support effective direct and timely collection and analysis of pandemic data, about both the progressIon of Covid-19 itself and the population response to public policy aimed at mitigating its progress. Refining policy and informing the judgement calls required to navigate the balance between lockdown and economic damage requires both accurate data and the ability to rapidly model and project the outcomes of multiple, 'What if?' scenarios. Current data intelligence systems are partial, fragmented, incomplete, lag reality and, in most cases can only surface what they have specifically been asked to look for. AI systems used to look for patterns are often constrained by the quality and range of data available to them. Existing hypothesis-driven models tend to look at single factors in isolation, and lack the flexibility to take into account multiple sources of mortality data or factors such as population mobility and behaviour, the impact of events such as Cheltenham races, sunny bank holiday weather or other regional and seasonal variations. This can only be addressed through a more holistic approach to data collection and integration. This project therefore uses an advanced data intelligence platform, udu, which has been used to integrate a wide range of data from multiple sources and of multiple types to create a unified, readily extensible and automatically updated data repository. This incorporates geospatial, pandemic, demographic and census data, population behavioural data from a number of sectors throughout the pandemic and makes it available for analysis. The project then uses a combination of mathematical modelling and inferential discovery to adjust for variations in data recording and then create daily projections of the course of the pandemic for all areas (currently in the UK), for up to two weeks ahead. These projections have been validated against the historical data for the pandemic and can be shown to provide predictions that are, for most areas, within +/-15% of the actual outcomes. Based on this breakthrough achievement, the project is now refining its analytic capabilities and projections by enabling segmentation by demographic, population behaviour and public policy (ie, which restrictions apply where, and when). An early outcome from its existing data will be a 'Smartcast' system to allow anyone to find out, on demand, what is happening and what restrictions apply in their area. The resulting system is intended to be capable of supporting direct exploration by human users and providing them with actionable predictions, as well as providing an interface (API) to allow other teams to access the datascape created, to then support third party analytics that further extend capability.