BI’s Article search uses Boolean search capabilities. If you are not familiar with these principles, here are some quick tips.
To search specifically for more than one word, put the search term in quotation marks. For example, “workers compensation”. This will limit your search to that combination of words.
To search for a combination of terms, use quotations and the & symbol. For example, “hurricane” & “loss”.
The data used to populate predictive models can come from either internal or external sources, and in either structured or unstructured formats.
While predictive models used by claims professionals have long relied more on internal, structured data, this is beginning to change, experts say.
Marty Ellingsworth, president of ISO Innovative Analytics, a unit of Jersey City, N.J.-based Insurance Services Office Inc., said advances in data mining methodologies have helped modelers glean new insights from unstructured data, such as the notes taken by a claims adjuster at the site of a claim.
“We've made a lot advances in the past three years, with new graph mining and text mining techniques,” Mr. Ellingsworth said. “Most of the best data is in adjuster's notes.”
These advances have radically altered the mix and amount of data that can go into claims models compared with just two or three years ago, Mr. Ellingsworth said.
For example, a claims modeler could pair these new insights from unstructured data with external financial information and mapping data in a novel manner to uncover a claims fraud ring operating in a given geography, Mr. Ellingsworth said.
“Claims that are suspicious may tend to cluster over time,” he said.
“Emerging fraud situations tend to generate cash flows, and we can now track how money and claims are pouring into these collusion networks at a faster rate than may be believable for a normal, growing business,” Mr. Ellingsworth said.
Using predictive modeling to augment claims management systems is expanding beyond insurers and third-party administrators, those familiar with the technology say.