A1037
Title: An analysis of mine-related insurance data using a compositional approach
Authors: Francesco Porro - Universita degli Studi di Genova (Italy) [presenting]
Abstract: One of the most critical issues faced by insurance companies is the continuous monitoring of the number of accidents experienced by their insured clients. An analysis of variables that can provide information on the forecasting of such events is therefore very worthy for insurance companies, especially if it is performed through innovative approaches and novel methodologies. A dataset provided by the US Mine Safety and Health Administration regarding the characteristics of a set of US mines, including the number of occurred accidents in the time range 2013-2016, is considered. Since most of the considered variables are either compositional or categorical, the analysis should be executed by means of the appropriate techniques. In particular, compositional data can be investigated by using the correct statistical tools in order to have reliable results. Following this approach, a compositional (CoDa) analysis is performed, taking into account that the relevant information conveyed by the compositional data is in the proportions among the parts and not in their absolute values or in their sum.