Measuring the relevance of factors in the occurrence of events

Abstract

A new way to compare the relevance of the different factors intervening in the occurrences of an event is presented and developed in this paper. The idea behind the method comes from cooperative game theory but the focus is slightly different because factors are not necessarily rational decision-makers and because the only data available are obtained by repetition of the event. The concept of relevance measure for a factor in a set of data is introduced, some significant examples are given and the main properties of relevance measures are defined and studied. One of these measures, the fair measure, is proved to have interesting properties which characterize it. Two real world situations, one about traffic accidents and the other one about mining accidents, both of them with real data, are used to show the use of relevance measures to compare factors in each one of these events.