Tag Archives: exposure management

EXPOSURE Magazine Snapshots: A New Way of Learning

This is a taster of an article published by RMS in the second edition of EXPOSURE magazine.  Click here and download your full copy now.

7 Apr 2017 - Machine Learning blog - Exposure banner image 720 x 168

 

In EXPOSURE magazine, we delved into the algorithmic depths of machine learning to better understand the data potential that it offers the insurance industry.  In the article, Peter Hahn, head of predictive analytics at Zurich North America illustrated how pattern recognition sits at the core of current machine learning. How do machines learn?  Peter compares it to how a child is taught to differentiate between similar animals; a machine would “learn” by viewing numerous different pictures of the animals, which are clearly tagged, again and again.

Hahn comments “Over time, the machine intuitively forms a pattern recognition that allows them to tell a tiger from, say, a leopard. You can’t predefine a set of rules to categorize every animal, but through pattern recognition you learn what the differences are.”

Hahn adds that pattern recognition is already a part of how underwriters assess a risk. “A decision-making process will obviously involve traditional, codified analytical processes, but it will also include sophisticated pattern recognition based on their experiences of similar companies operating in similar fields with similar constraints. They essentially know what this type of risk ‘looks like’ intuitively.”

The Potential of Machine Learning

EXPOSURE magazine asked Christos Mitas, vice president of model development at RMS, on how he sees machine learning being used.  Mitas opened the discussion saying “We are now operating in a world where that data is expanding exponentially, and machine learning is one tool that will help us to harness that.”

Here are three areas where Mitas believes machine learning will make an impact:

Cyber Risk Modeling: Mitas adds “Where machine learning can play an important role here is in helping us tackle the complexity of this risk. Being able to collect and digest more effectively the immense volumes of data which have been harvested from numerous online sources and datasets will yield a significant advantage.”

Image Processing: “With developments in machine learning, for example, we might be able to introduce new data sources into our processing capabilities and make it a faster and more automated data management process to access images in the aftermath of a disaster. Further, we might be able to apply machine learning algorithms to analyze building damage post event to support speedier loss assessment processes.”

Natural Language Processing: “Advances here could also help tremendously in claims processing and exposure management,” Mitas adds, “where you have to consume reams of reports, images and facts rather than structured data. That is where algorithms can really deliver a different scale of potential solutions.”

For the full article and more insight for the insurance industry, click here and download your full copy of EXPOSURE magazine now.

For more information on RMS(one)®, a big data and analytics platform built from the ground-up for the insurance industry, and solutions such as Risk Modeler and Exposure Manager, please click here.

Fire Weather

Fires can start at all times and places, but how a fire spreads is principally down to the weather.

This week, 350 years ago, the fire at Thomas Farriner’s bakery on Pudding Lane, a small alleyway running down to the river from the City of London, broke out at the quietest time of the week, around 1am on Sunday morning September 2, 1666. London had been experiencing a drought and the thatched roofs of the houses were tinder dry. At 4 am the Lord Mayor, roused from his sleep, decided the blaze was easily manageable. It was already too late, however. By 7am the roofs of some 300 houses were burning and fanned by strong easterly winds the fire was spreading fast towards the west. Within three days the fire had consumed 13,000 houses and left 70,000 homeless.

In the city’s reconstruction only brick and tiles houses were permitted, severely reducing the potential for repeat conflagrations. Within a few years there were the first fire insurers, growing their business as fear outran the risk.

Yet big city fires had by no means gone away and the wooden cities of northern Europe were primed to burn. The 1728 Copenhagen fire destroyed 28% of the city while the 1795 fire left 6000 homeless. A quarter of the city of Helsinki burned down in November 1808. The 1842 fire that destroyed Hamburg left 20,000 homeless. The center of the city of Bergen Norway burnt down in 1855 and then again in January 1916.

Wind and fire

By the start of the 20th Century, improvements in fire-fighting had reduced the chance that a great city fire took hold, but not if there were strong winds, like the 1916 Bergen, Norway fire, which broke out in the middle of an intense windstorm with hurricane force gusts. In February 1941 the fire that burnt out the historic center of Santander on the coast of northern Spain was driven by an intense windstorm: equivalent to the 1987 October storm in the U.K. And then there is the firestorm that destroyed Yokohama and Tokyo after the 1923 earthquake, driven by 50 miles per hour winds on the outer edge of a typhoon in which, over a few hours, an estimated 140,000 died.

Wind and fire in the wooden city are a deadly combination. Above a certain wind speed, the fire becomes an uncontrollable firestorm. The 1991 Oakland Hills fire flared up late morning also on a Sunday and then surged out of the mountains into the city, driven by hot dry 60 miles per hour Diablo Winds from the east, jumping an 8 lane highway and overwhelming the ability of the fire crews to hold the line, until the wind eventually turned and the fire blew back over its own embers.  The fire consumed 2800 houses, spreading so fast that 25 died. On February 7, 2009 a strong northwesterly wind drew baking air out of Australia’s interior and fires took off across the state of Victoria. Fallen power cables sparked a fire whose embers, blown by 60 miles per hour winds, flashed from one woodland to another, overwhelming several small towns so fast that 173 died before they could escape.

Most recently we have seen fire storms in Canada. Again there is nothing new about the phenomenon; the Matheson fires in 1919 destroyed 49 Ontario towns and killed 244 people in a fire front that extended 60km wide. It was a firestorm fanned by gale force winds, that destroyed one third of the city of Slave Lake, Alberta, in 2011 and it is fortunate only that the roads were broad and straight to allow people to escape the fires that raged into Fort McMurray in summer 2016.

There is no remedy for a firestorm blown on gale-force winds. And wooden property close to drought ridden forests are at very high risk, such as those from South Lake Tahoe to Berkeley in California and in New Zealand, from Canberra to Christchurch. Which is why urban fire needs to stay on the agenda of catastrophe risk management. A wind driven conflagration can blow deep into any timber city, and insurers need to manage their exposure concentrations.