ANALISIS CADANGAN KLAIM MENGGUNAKAN GENERALIZED LINEAR MODEL DAN HIERARCHICAL GENERALIZED LINEAR MODEL

LUTVIARINI LATIFAH, Dr. Adhitya Ronnie Effendie, M.Sc.

2016 | Tesis | S2 Matematika

Cadangan klaim merupakan salah satu permasalahan penting pada asuransi kerugian. Pada asuransi kerugian, klaim terhubung dengan periode dibayarnya premi. Pemodelan seperti ini dapat menggunakan kerangka run-off triangle. Metode Chain-Ladder merupakan salah satu metode prediksi cadangan klaim yang bekerja pada skema run-off triangle kumulatif. Metode Chain-Ladder memberikan hasil cadangan klaim yang sama dengan model overdispersed poisson dari Generalized Linear Model. Kelebihannya model overdispersed poisson memberikan ukuran kesalahan prediksi. Namun model overdispersed poisson mengasumsikan data independen tanpa memperhitungkan karakteristik periode kejadian yang mempengaruhi prediksi dan kesalahan prediksi model. Model overdispersed poisson-gamma dari Hierarchical Generalized Linear Model mempertimbangkan karakteristik periode kejadian yang mempengaruhi prediksi cadangan klaim. Selain itu, model overdispersed poisson-gamma memberikan nilai kesalahan prediksi yang lebih kecil dibandingkan dengan model overdispersed poisson.

Claims reserving is one of important issues in general insurance. In general insurance, claims are connected to the payment premium-due. When modelling this situation, run-off triangle frameworks can be used. In practice, the most popular statistical method in the claims reserving literature is Chain-Ladder method that works in cumulative run-off triangle frameworks. Chain-Ladder method gives the same reserve as Overdispersed Poisson model, from Generalized Linear Model. In addition, the prediction error of Overdispersed Poisson model is measurable. However, the model assumes that independent data without taking into account the characteristics of the accident period that affect to the prediction error. Overdispersed Poisson-Gamma model from Hierarchical Generalized Linear Model considers the characteristics of accident period that affect claims reserve prediction. In addition, the prediction error value of this model is smaller than the Overdispersed Poisson model.

Kata Kunci : Asuransi Kerugian, Cadangan Klaim, Skema Run-off Triangle Data, Generalized Linear Model, Hierarchical Generalized Linear Model, Kesalahan Prediksi.