Czech Insurers' Bureau
The Czech Insurers' Bureau (ČKP) is a professional organization of insurers authorized to provide motor third-party liability insurance in the Czech Republic. In addition to its other activities, which are likewise established by law, the ČKP cooperates with state authorities in matters relating to liability insurance, maintains records and statistics for the purposes of liability insurance, and at the same time takes part in preventing accidents in road traffic and in preventing insurance fraud in insurance related to the operation of vehicles.
Starting situation
The ČKP had already been using Statistica tools for a long time — specifically the Statistica Analyst and Statistica Data Scientist modules. The latter software is also used to analyze data on traffic accidents and „motor third-party liability“ insurance claims. With the aim of broadening the overall scope of the analyses performed, a demand arose for a tool able to work with unstructured texts — Text Mining. Given its long-standing satisfaction with the Statistica software and the requirement that the extension module be compatible with the already-used analytical systems mentioned above, the ČKP sought an analytical tool for Text Mining.
The ČKP's requirements
The need to analyze unstructured text arises in practice, for example, in situations where it is necessary to more precisely quantify the information required to classify claims as serious and less serious. The share of personal-injury claims in the total liabilities of motor third-party liability insurance (MTPL) is, in fact, continually rising.
The associated most serious annuity claims generate enormous costs for insurers, amounting to as much as 50% of the volume of all personal-injury claims.
RNDr. Petr Jedlička, Ph.D., Head of the Actuarial and Analytics Department at SUPIN, the service organization of the ČKP and the Czech Insurance Association (ČAP)
The ČKP expects that, by combining the application of the statistical techniques of Text Mining and Data Mining, it will be possible to expand the risk indicators that predict the occurrence of serious personal-injury claims already at an early stage of investigating
the case. This will allow more accurate estimates of the total amount of claims.
Implementation process
Owing to the maturity of the product, the entire software implementation took, in effect, a single day. From the ČKP's perspective, the initial configuration, the implementation itself, the brief training, and the removal of implementation obstacles all took place in a very short time and at the standard (i.e. high) quality.
Final situation
The implementation proceeded according to the original plan, and the solution meets all of the ČKP's requirements. As expected, the combination of the Statistica Analyst, Statistica Data Scientist, and Text Mining module tools makes it possible to expand the usable information drawn from the analyzable data. Specifically, the two techniques work as follows: the role of the Text Miner is to extract representative keywords from unstructured textual information (a description of how the accident occurred), which are then processed into standard data fields. From the data fields prepared in this way, using the Data Scientist tool and its Data Mining module, a more precise estimate of the personal-injury claim amount is obtained for specific accident scenarios that, without the analysis of unstructured text, would not be detectable or statistically evaluable. The achieved goal of refining the estimate of the serious personal-injury claim amount will make it possible to better estimate the total MTPL claim payouts, which leads to better risk management within the most serious segment of MTPL claims. Ultimately, this results in accurate resource planning in the insurance sector.