Predictive analytics: visions of tomorrow
Valérie Bécaert Valérie Bécaert
November 20 5 min

Predictive analytics: visions of tomorrow

Throughout history, human beings have been captivated by the idea of seeing into the future. From the Oracle of Delphi in Ancient Greece to the speculative science fiction of Isaac Asimov and Stanley Kubrick, it’s an obsession we’ve always had. Predictive analytics is the latest expression of this fundamental human desire to get a glimpse of tomorrow before we get there.

In this article, we’ll explore the basics of this fascinating field, how it fits into the wider landscape of machine learning and artificial intelligence and the real-world applications that continue to change the way we live our lives.

A world of data to understand

The term “data” has high-tech connotations, but in essence it just means information. And so, data analysis really isn’t new. It’s something we’ve always done, as a species — gathering information, thinking about it and attempting to draw conclusions. Except that now we have access to an unprecedented abundance of information — via developments such as the Internet of Things — and much more sophisticated ways of processing it.

We’re dealing with higher volumes of information than ever before and capturing it in greater depth and complexity. A human can look at all this data and attempt to glean meaning from it, but there’s only so much a person can handle. Computers led to a revolution in data analysis, but they can only do exactly what they’re programmed to do. To really unlock the value present in data, we need to know what we don’t know. And that’s where AI comes in.

Data analysis by the digital native

Using data, AI can be used to create complex models of what may happen in the future: simulating potential events by using past experience as a basis, deciding what’s likely to be the best path ahead. University of Toronto professor Ajay Agrawal suggests that AI is “machine prediction,” in the same way that the earliest computers were performing “machine arithmetic.”

Machine learning now allows an unprecedented level of insight in data analysis and more realistic forecasts. Instead of merely collating and presenting the data for us to analyze, AI can make predictions and come to decisions without direct human intervention — and, if those outcomes are wrong, an AI system can be built with the ability to correct itself.

In some cases, more traditional statistical approaches are still better than AI models based on deep learning models, but we are only beginning to harness the power of AI. As more and more data is fed into self-correcting AI models, they are becoming more and more powerful.

There are still challenges for AI prediction, not least because AI systems are good at decision-making but bad at explanations. Still, the power of learning algorithms is a true paradigm shift in data analysis.

We have the tools now to make more accurate predictions than ever before, and the applications are limitless. In his book Prediction Machines: The Simple Economics of Artificial Intelligence, professor Agrawal suggests that, just as the invention of the computer led to the end of film cameras, we’re still too early to see the true impacts of machine prediction.

Businesses founded on anticipating the future

From high-level scientific research to market forecasts, predictive analytics powered by AI has changed how we engage with tomorrow’s possibilities. Today’s insurance industry is a good example: it’s a field that’s fundamentally concerned with risk and being prepared for future events.

According to a report from Willis Towers Watson, the third-largest insurance broker in the world, 54% of insurers utilize predictive models to collect data regarding their customers and their businesses.

Underwriting’s core business model is to collect more income (e.g. insurance premiums) from customers than they’re paying out in losses. To the layman, that may seem obvious and a simple matter, but there are many issues at play that decide success or failure. Pricing is key, and complex calculations are required in order to both remain competitive and provide the best balance in terms of incoming profits and outgoing payments (should they be needed).

To optimize this, AI models can decide the pricing structures and rates for underwriters, using historical data and modelling future probabilities of losses, customer behaviour, market changes, and other factors. They can also use analytics to detect possible current or future instances of fraud, accessing wider federated or governed data sources to check a customer’s past and see if they’ve been referred for an investigation, had prior claims denied, or are associated with suspicious people or organizations. Now AI can play detective as well as analyst.

Business success is increasingly determined by an organization’s ability to harness its technology and data to its fullest potential. And it’s only going to become more so in the future as AI unlocks new insights from data.