Institute of Mathematics for Industry

Seminar



List All(1084) Today and tomorrow's seminars(2)

Applications of Predictive Modeling in Actuarial Science


Hold Date 2019-10-09 16:45~2019-10-09 17:45

Place IMI Auditorium (W1-D-413), West Zone 1, Ito campus, Kyushu University

Object person researchers and graduate students 

Speaker Hirokazu IWASAWA (Kobo Iwahiro)

The IMI Colloquium in October
 
Date : Wednesday, 9 October 2019
          16:45-17:45
Place : IMI Auditorium (W1-D-413) (4F, West Zone 1, Kyushu University)
 
Speaker : Mr. Hirokazu IWASAWA (Kobo Iwahiro)
 
Title :  Applications of Predictive Modeling in Actuarial Science
 
Abstract :
In this talk, predictive modeling for dealing with risks will be introduced from the viewpoint of practitioners, especially from that of actuaries.

Predictive modeling can be defined as "the process of developing a mathematical tool or model that generates an accurate prediction" while this concept itself is not quite new. Its importance, however, is increasing rapidly in the era of Big Data. Thus, today's actuaries are required to master the basics of predictive modeling as an international standard. The speaker will present an outline of what are the "basics" of predictive modeling in this context.

Risks will be also focused in this talk. Here they refer to those that are "uncertain" and "wanted to be avoided," and are typically handled directly by insurance companies and other financial institutions. Practitioners in the field need to have predictive models that maintain simplicity and perform high accuracy in dealing with data with high uncertainty. However, most predictive modeling methods widely introduced today are those that perform well for tasks of pattern recognition with low uncertainty. In fact, as there has not been adequate assortment of models that satisfy the practitioners working with risk, it is still necessary to develop models starting from rather basic ones. There will be, in this talk, a brief introduction of a basic but brand-new model, called AGLM (see: https://github.com/kkondo1981/aglm), that the speaker is involved in developing.
 


IMI Colloquium Report