A Qualitative Method for Risk Preference of Insurance Companies

The risk evaluation problem of insurance companies often has two prominent characteristics: one is that there are multiple risk attributes with the different weights, the other is that the risk evaluations are the natural words or qualitative descriptions. The latter characteristic provides a big challenge for the qualitative aggregation method. This paper presents a new method to deal with this problem. The start point of this method is a finite set of basic qualitative descriptions or words and a semantic similarity relation on these words. For any complex linguistic expression for evaluating the underlying insurance company we have a semantic similarity distribution on the the basic words, from which a risk score can be derived. by aggregating all risk scores for the underlying insurance company we can get the final risk score which synthesizes all risk attributes. The main advantages of this aggregation method are that the complex qualitative evaluations having natural word forms can be operated appropriately, and the semantic overlapping among qualitative evaluations is also considered in a natural way.