DATA-MINING HELPS INSURERS DEAL WITH SUBSIDENCE CLAIMS

Issue date: 23 May 2001


Subsidence is a major headache not only for the householder, but also for the buildings insurance company who has to pick up any bills for remedial work. Faced with the problem of predicting costs of repairs, and therefore the level of future insurance premiums, Lloyds TSB Insurance, the number one distributor of new home insurance policies*, turned to the University of the West of England for a solution.

At present, the company's initial estimates for subsidence claims are based on survey reports, which can prove inaccurate. These discrepancies have to be managed using a probability weighting technique. In addition, claims sometimes reopen and represent further costs to the insurer. Lloyds TSB Insurance wanted to take this opportunity to improve estimates of subsidence repairs, using historical data. Staff from UWE's faculty of Computer Studies and Mathematics proposed the development of new computer modelling techniques for discovering patterns in (or mining) data in this way.

" Evolutionary computing techniques have been used before for data mining, as they are able to search large problem spaces in parallel in a relatively unsupervised manner," said Dr Larry Bull. "In this project we proposed evolving a form of rule-based system which has been used to generalise successfully in complex situations such as this."

"We considered a number of universities for this important research," said Paul
Laughlin, Senior Risk Manager from Lloyds TSB Insurance. "UWE provided the most convincing and innovative proposal. Our work with UWE in the past has proven the depth of their expertise in artificial intelligence and modelling. We look forward to working together on improving our understanding in new areas."

Similar techniques will also be applied to dealing with claims due to unemployment against the company's personal loan protection insurance. A strong correlation had been noticed between the frequency of claims and changes in national unemployment levels. The same project aims to improve the accuracy of the company's current performance in predicting the extent of these claims.

-ENDS-

Editor's Notes
* According to independent research published by NOP Research Group, September 2000

FFI: Jane Kelly or Mary Price, Press Officers
BRISTOL UWE
Tel: 0117 3442208; fax: 0117 976 3912;
E-mail: Jane.Kelly@uwe.ac.uk or Mary.Price@uwe.ac.uk


Back to top