The where, when and why of data

When we think of data, typically what comes to mind is raw information: long strings of digits that can be mined for this or that micro-insight which, eventually and in aggregate, paints an abstract picture of some sort.

Such data does exist, but there is also more immediately representational data.

Spatial data, for example, is a crucial area for analytics, and seeks to paint a much more representational picture.

“A map is a representation of reality that has its basis in data, that data having x, y and z coordinates, and possibly time as well,” Eamonn Doyle, chief technology officer, Esri Ireland, said.

As a mapping and spatial data systems specialist, Esri Ireland has become the leading actor in the technology used to represent the country.

“There are generally four dimensions to the data we are dealing with, but always at least two and usually at least three,” he said.

Applications can be sophisticated. In recent years, Esri Ireland has found it is using space-time data mining for pattern recognition.

“It’s, very much, not just hotspot mapping. Instead, it’s trying to find intersections of time and space that are either the norm or else outliers,” Doyle said.

Applications in areas such as transport and traffic management are obvious.

“It’s particularly important in terms of things like road traffic accidents. You can look at where they are happening, when they are happening, and then start to ask questions as to why they are happening. Are they happening during weather events, for example.

“Likewise, you might be looking at a fleet of buses and asking if they are arriving on time and if not, why not,” he said.

Clients interested in the matrix of spatial and temporal information tend, at the moment, to be in the public sector, and oftentimes they are seeking to make sense of data that they already have.

It's not necessarily always work in the realm of ‘big data’, Doyle said, it could be things like location analytics where they can rank one location according to set characteristics and compare it to others.

“They might want to narrow down 20 or 30 potential locations for an FDI investment: how big are the premises, how well serviced is it, and then do site selection on that,” he said.

Another use is the provision of public services.

“So, for instance, with the influx of Ukrainian refugees, government is interested in where those refugees are going to be in the short, medium and long term, and what capacity exist at those locations in terms of education, health and all of the other services that the state provides, as well as what the impact will be on existing provision of services by increasing the population in an area,” he said.

“The corollary of that is: ‘how do we increase services’,” he said.

As a result, there is a major predictive element to spatial and spatio-temporal data analysis. A lot of state services depend not just on where service users are, but also what age cohort they are in. For example, if you know someone’s age now, you also know what age people will be in five years’ time.

“This is something we’ve been doing with the department of education for some time. They would have predictive analysis in provision of school places, for example. There are also various projects within the department of housing, asking what services are available for older people in a given area,” Doyle said.

The point of all of this data is not just simply to create fodder for processing. Instead, it is designed to assist and augment decision-making – decisions made by people.

“Quite often, the best intelligence is the intelligence between your ears. You don't always need AI, but sometimes you need it where there is a volume of data that needs to be worked through,” Doyle said.

Esri Ireland’s applications for AI include object detection.

“We do a lot of work with Ordnance Survey Ireland, working with aerial images, and feature extraction is a good way of looking at the built environment,” he said.

For example, AI can pick up changes in the built environment: it can see where a new housing estate has been built and there are, for example, 243 new houses in it, covering a certain area.

“Feature extraction is very useful and it’s a way of finding that sweet spot where the technology can do something for you,” Doyle said.