Speaker(s): Stefan Wolfgang Wetzel (ALTE LEIPZIGER Versicherung AG)
A classic tariff in private accident insurance in Germany is usually split up into in two categories of professions: professions without and with physical activity ("Gefährdungsklasse" A and B). In recent years more categories have been introduced in the market, mainly by splitting the more expensive category B in two or more parts. Such classifications usually rely on the judgment of specialists.
In order to achieve a more detailed and risk-oriented price, we propose a new ansatz of classification. Instead of relying on specialists for grouping professions, we use crowd-intelligence to create a metric multi-dimensional space of professions. In general the number of dimensions for such a space would be N-1, with N being the total number of considered professions. This number is usually much larger than 1000. Thus the results of our attempt would be too complex and unusable for real world tariff calculation. To solve this problem, we introduce a model that allows us to choose the number of dimensions in advance. The coordinates of all professions in this space are then fitted using a proper metric. The newly found coordinates are used to regionalize the accident tariff with respect of this multi-dimensional space of professions. The result is clustered for the use in the final tariff.