Cities should focus on ensuring that their data is usable as a key first step in establishing and supporting smart city systems, says analytics multinational SAS advanced analytics and artificial intelligence expert Kelly Lu.

The preparation of data prior to use is also an effective way for cities to evaluate the readiness of their own structures and personnel to manage eventual smart city systems. This will assist in raising the importance of data science, and in cleaning and organising historical data as necessary elements to support such systems.

“WiFi hot spots and smart traffic are part of the hype around smart cities. Most smart city systems will typically be invisible to citizens. They are about data and the analytics that can be performed using this data to provide valuable outputs,” she adds.

For example, sensor networks can allow for the accurate monitoring and billing of electricity use, support safety and security, and provide leak detection long before there is a complaint.

“All the systems will be working in the background to ensure that city dwellers have a better standard of living. The systems must be smart, convenient and inclusive. We need to focus on what counts and makes the biggest positive change,” says Lu.

Cities are natural starting points, owing to the density of the population and the relative wealth of cities, compared with the hinterland. More than half of people worldwide live in cities, and urban areas produce about 70% of nations’ wealth.

While South Africa does not have smart city infrastructure in place – such as sensors and camera networks – pilot projects can be deployed in urban settings and cover sufficient people to provide a good baseline without a massive investment in infrastructure and sensors, she adds.

“This approach will enable cities to understand the infrastructure required and can be expanded if successful. We should be testing the concepts to determine what works and then assess the challenges and obstacles,” advises Lu.

Once the sensors are in place and monitoring consumption, detecting leaks and storing the data, they will create historical data and drive descriptive analytics, which looks for peaks, troughs and trends to provide insights.

“This is where we can understand how to use what we have better and how to use the systems better. Once we understand the causes and effects better, we can then understand when to do maintenance and when to develop or expand infrastructure,” she explains.

Using the historical and weighted data, the projects can then predict and forecast the areas that are likely to boom and require additional infrastructure and utilities, as well as predict the extent of demand. Similarly, the structured data will also enable, for example, a water utility to predict when a burst will occur by looking at the historical data and the anomalies that are associated with a burst and which occurred prior to a burst.

A particular barrier, however, is that retrofitting sensors is more difficult than installing them during infrastructure development, and smart city systems are competing for capital with more pressing issues in a developing country.

“To overcome this barrier, we have to try to use what is there already and prepare for when additional data sources can be added.”

Similarly, deploying analytics engines close to the edge of a network is ideal, but impractical with existing infrastructure. The data feeds would have to be sent to a database and the analytics can then be run on top of this.

Lu warns that any analytics and smart city systems cannot be a one-size-fits-all undertaking.

“Each department needs its own model, specifically models that address the problems at hand. Simultaneously, there is a need to ensure that each analytical model is in place only while it remains relevant and suitable.”

This requires an understanding of the lifecycle of analytical models and data science knowledge.

“There is a shortage of data scientists across industries. Many young data scientists also have shorter tenures at companies, as demand enables them to pursue more exciting opportunities. Tenures tend to be shorter than a year.”

In addition to this skills dearth, the specialised nature of the business problems cannot be solved only by data scientists, as they may not know the relevance or importance of various variables and, consequently, may not be able to identify the reasons a smart city analytical model is decaying.

“A simple fix may be to present young data scientists with exciting projects that they can tackle,” says Lu.