Predictive modelling: managing a modern city

Recently, Chicago has implemented a policy to include predictive models in the tools available to prevent a variety of potential threats, from infestations to poisoning.

Big cities with high population densities and industrial development face constant problems that can be difficult to manage accurately and within the desirable timeline. How is it possible to prevent occasional food intoxication, lead poisoning from water pipes, cockroach infestations, landslides, or gas leaks, in a city of 2.7 million without knowing where to search?

Tons of data are collected everyday in various ways, be it through company surveys, geological studies, biological analyses, chemical tests, police reports, university studies, quality inspections, or as preparation for project approvals. It can be argued that they serve their purpose, but could they have a further use? The answer is YES. And data scientists have long known this.

Data scientists are people who analyse big chunks of data for a living. Sounds nerdy and slightly boring, right? Wrong. What with globalisation and the ‘technology generation’, it would be criminal not to use all the resources we have to maximise the information we can get from research. Through the sharing of data and its analysis by skilled professionals, steps can be taken towards a greater easiness of life.

Big data prediction models can be applied to historical data in order to convert it into predictions that can be used to know the likelihood of certain events occurring. This way, local authorities can build a decent picture of where to look for potential threats, meaning that they can act faster and prevent major catastrophes.

The government of Chicago is committed to applying predictive analytics in public health management, following the impressive results of pilot tests. The latest project, still in evaluation and preliminary tests, aims to analyse home inspection records and census data to predict the likelihood of buildings causing lead poisoning in children.

The use of big data in public health control is not, however, completely new. For instance, a method that links municipal complaints about rats to the general conditions of the area that may attract rodents – be it excessive dirt, sewage or water leakage – has already been implemented. Health agents can then target sanitation teams to potential areas, leading to 15% fewer requests for rodent control in the past year.

Although models are only capable of predicting certain types of event, and rely heavily on previous records, it is still a better approach than none. Other cities, such as New York and Dublin, have already adopted predictive analytics for various purposes, from fire risk to crime rates.

Predictive modelling is growing fast and definitely seems to be a very promising tool for government agencies. Exciting times for data scientists, I would say!