Quantcast
Last updated on April 19, 2014 at 9:20 EDT

New Big Data Platform Increases Reliability of Property Catastrophe Pricing

October 25, 2012

GREENWICH, Conn., Oct. 25, 2012 /PRNewswire/ — To meet the demand for more accurate Expected Loss estimates from catastrophe models, Datum Companies today announced Eurus®, a cutting edge catastrophic weather predictive platform. Nearly two years in the making and powered by a combination of Big Data analytical technologies–including Wavelets, Support Vector Machines, Principal Components Analysis and Neural Networks–the Eurus® platform understands complex, non-linear relationships between millions of historical weather feature data points using machine learning to calculate catastrophic event probabilities. Fundamentally different than traditional catastrophe models, Datum’s innovation delivers enhanced results.

Explains Sean Bourgeois, Datum Companies President & CEO, “The approach that property catastrophe modeling companies currently use to calculate the Hazard component of their models isn’t sufficiently predictive, so legacy models shouldn’t be used in isolation if industry participants want to accurately price catastrophic risk. As we all know, the property catastrophe models that insurance and reinsurance companies have been using for years have poor track records when it comes to estimating medium- to long- term Expected Loss, so the industry instead applies simple multiplication factors to Expected Loss results to estimate catastrophic risk. Additionally, due to the mechanics of these models, frequent revisions have been necessary which reset Probable Maximum Loss calculations. Inexact? You bet. We developed Eurus® to measurably boost predictive accuracy.”

Because the new Eurus® platform is based on a global grid, it can address catastrophic weather risk anywhere in the world, including presently unmodeled regions. Eurus® platform capabilities exceed traditional models, providing consistently accurate predictive results as far as one year into the future. And Datum’s platform is less susceptible to change following major catastrophic events while being less sensitive to weather phenomena like El Nino/La Nina. As the platform continually ‘learns’ incrementally across a massive data set, periodic model rewrites are no longer required. The new platform delivers more stable and reliable results across time.

Datum’s world class technology infrastructure includes a massive NoSQL Cassandra data warehouse. Its machine learning models are powered by over 4 billion historical weather parameter readings representing more than 10 Terabytes of data. Datum deploys Elastic MapReduce across a Hadoop cluster, providing unparalleled computational power and efficiency to enable its predictive modeling capabilities.

Eurus® can be used as a stand-alone tool or incorporated with legacy model results to improve predictive estimates. Eurus® can ingest multiple model result files and reprocess event catalogs to revise probabilities-per-event and recalculate monetary loss estimates to generate secondary Exceedance Probability curves and comparative Expected Loss calculations. As a result, the Eurus® platform lets industry participants better understand and assess catastrophe risk pricing and PML calculations.

Access to the Eurus® platform is available on a subscription basis. The platform’s applications run entirely in the cloud on Datum’s proprietary technology infrastructure. There is no software to install or upgrade. Datum handles all data processing and delivers actionable results.

About Datum

Datum Companies is a predictive analytics technology company serving the insurance and reinsurance industry. The Datum team includes data scientists, physicists, engineers, mathematicians, software developers and insurance industry experts, all working together to develop applications that solve important industry problems using Big Data and cutting edge technology. Datum Companies is based in Greenwich, CT and is privately held. For more information, visit www.datumcos.com.

SOURCE Datum Companies


Source: PR Newswire