DETERMINATION OF EARTHQUAKE INSURANCE BUSINESS CLAIM RESERVE BY APPLYING NEURO FUZZY INFERENCE SYSTEM: GRADIENT-DESCENT METHOD

• Main Author: Senjaya, Wendy
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39296
• Institution: Institut Teknologi Bandung

An insurance company need to predict its claims reserve. The Basic Chain Ladder method is the most common method used to determine the outstanding claims liability of a long-tail insurance business because this method is simple and distribution free. If the Basic Chain Ladder method is used to determine the outstanding claims liability of an earthquake insurance business, then often the characteristics of each claim are not taken into account. In this Final Project, a Neuro Fuzzy Inference System method which takes into account the characteristics of an earthquake insurance claims data is constructed. The characteristics used are: moment magnitude; depth; latitude and longitude. The earthquake insurance claims data is grouped into several clusters. Then, for each cluster, a run-off triangle is formed and its development factors are obtained based on the Basic Chain Ladder method. Initially, it is assumed that the clustering is based on the earthquakes’ cresta zones. However, the results of the analysis showed that the clustering is based on the depths of the earthquakes’ hypocenters.