Machine learning & experience analysis

In this presentation we  review  existing  modelling  approaches  for  analysing  claims  experience  in  the  presence of reporting delays, reviewing the formulation of mortality incidence models such as GLMs. We then show how these approaches have traditionally been adjusted for late reporting of  claims  using  either  the  IBNR  approach  or  the  more  recent  EBNER  approach.  We  then  go  on  to  introduce  a  new  model  formulation  that  combines  a  model  for  late  reported  claims  with  a  model  for  mortality  incidence  into  a  single  model  formulation.  We  then  illustrate  the  use  and  performance  of  the  traditional  and  the  combined  model  formulations  on  data  from  a  multinational  reinsurer.  We  show  how  GLMs,  lasso  regression,  gradient  boosted  trees  and  deep  learning  can  be  applied  to  the  new  formulation  to  produce  results  of  superior  accuracy  compared to the traditional approaches