Χρήση τεχνικών εξόρυξης δεδομένων στον αναλογισμό
Data mining techniques in actuary and insurance
SubjectΕξόρυξη δεδομένων ; Ασφαλιστικές εταιρείες ; Data mining ; Mathematical statistics ; Insurance companies
Almost 80% of the data utilized on a daily basis all over the world are unstructured. Modern decision support systems are based on information extracted from structured data, thereof neglecting unstructured data that can also provide significant information. Therefore there is a need for new dynamic tools that will help to transform unstructured data into structured information which, combined with the information extracted from structured data will help information consumers make better decisions. A new field of science, named Text Mining promises to fill this gap. text mining is a combination of different sciences such as statistics, machine learning, information theory and computational procedures. In this thesis we examine two text mining applications which are of major inportance for the insurance sector. In the first text we illustrate how can be classified into groups based on text subjects or according to other attributes (common words, frequency of words etc.). This application is very useful for document analysis and useful information extraction. In the second we present how forecasting techniques are applied in text mining to predict, for example if a document contains positive or negative comments or if an insurance claim is fraud or not. At the same time we present the most important concepts and methods used by classification and prediction techniques in text mining. The aim of this thesis is to present how useful text mining tools can be to an insurance company and elucidate an easy and simple way to extract important information from unstructured or semi-structured documents.