Estimasi Cadangan Klaim Menggunakan Metode Generalized Linear Model Tweedie Compound Poisson Dengan Estimasi Parameter Indeks Melalui Extended Quasi-Likelihood

Sari, Dwiani Novita (2026) Estimasi Cadangan Klaim Menggunakan Metode Generalized Linear Model Tweedie Compound Poisson Dengan Estimasi Parameter Indeks Melalui Extended Quasi-Likelihood. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221054-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Estimasi cadangan klaim, khususnya Incurred But Not Reported (IBNR), merupakan aspek penting dalam menjaga solvabilitas dan stabilitas finansial perusahaan asuransi. Penelitian ini bertujuan untuk mengestimasi cadangan klaim menggunakan metode Generalized Linear Model (GLM) dengan pendekatan distribusi Tweedie Compound Poisson (TCP) serta mengestimasi parameter indeks menggunakan metode Extended Quasi-Likelihood (EQL). Distribusi TCP dipilih karena mampu memodelkan data klaim yang memiliki proporsi nilai nol tinggi (zero-inflated) dan variabilitas heterogen. Data yang digunakan berupa data sekunder paid claims asuransi kendaraan bermotor PT Asuransi XYZ periode Agustus 2022 hingga Agustus 2025 yang disusun dalam bentuk run-off triangle inkremental bulanan. Estimasi parameter model dilakukan menggunakan metode Maximum Likelihood Estimation (MLE) dengan algoritma Iteratively Reweighted Least Squares (IWLS). Estimasi parameter indeks dilakukan dengan mengevaluasi 99.999 kandidat nilai parameter pada rentang 1 hingga 2 menggunakan pendekatan EQL untuk memperoleh nilai parameter indeks yang paling sesuai dengan karakteristik data. Evaluasi model dilakukan menggunakan Root Mean Square Error of Prediction (RMSEP) dan confidence interval untuk mengukur tingkat ketidakpastian estimasi. Hasil penelitian menunjukkan bahwa nilai parameter indeks optimum yang diperoleh melalui metode EQL adalah sebesar 1,42586, yang menunjukkan bahwa data mengikuti distribusi Tweedie Compound Poisson. Berdasarkan model GLM TCP yang dibentuk, diperoleh total estimasi cadangan klaim sebesar Rp5.831.472.598 dan nilai RMSEP yang dihasilkan sebesar Rp1.262.210.807. Selain itu, diperoleh interval kepercayaan 95% dengan batas bawah sebesar Rp3.357.539.417 dan batas atas sebesar Rp8.305.405.779. Hasil penelitian ini menunjukkan bahwa pendekatan GLM Tweedie Compound Poisson dengan estimasi parameter indeks melalui EQL mampu memberikan estimasi cadangan klaim yang fleksibel dan representatif terhadap karakteristik data klaim yang kompleks. ======================================================================================================================================
Claim reserve estimation, particularly Incurred But Not Reported (IBNR) claims, is an important aspect in maintaining the solvency and financial stability of insurance companies. This study aims to estimate claim reserves using the Generalized Linear Model (GLM) with the Tweedie Compound Poisson (TCP) distribution approach and to estimate the index parameter using the Extended Quasi-Likelihood (EQL) method. The TCP distribution was selected because it is capable of modeling claim data with a high proportion of zero values (zero-inflated) and heterogeneous variability. The data used in this study were secondary data consisting of paid claims from motor vehicle insurance at PT Asuransi XYZ for the period August 2022 to August 2025, arranged in the form of a monthly incremental run-off triangle. Model parameter estimation was carried out using the Maximum Likelihood Estimation (MLE) method with the Iteratively Reweighted Least Squares (IWLS) algorithm. The index parameter estimation was conducted by evaluating 99,999 candidate parameter values within the range of 1 to 2 using the EQL approach to obtain the index parameter value that best fits the characteristics of the data. Model evaluation was performed using the Root Mean Square Error of Prediction (RMSEP) and confidence interval to measure the level of estimation uncertainty.The results showed that the optimum index parameter value obtained through the EQL method was 1.42586, indicating that the data followed a Tweedie Compound Poisson distribution. Based on the constructed GLM TCP model, the total estimated claim reserve obtained was IDR 5,831,472,598 and the resulting RMSEP value was IDR 1,262,210,807. In addition, the 95% confidence interval obtained had a lower bound of IDR 3,357,539,417 and an upper bound of IDR 8,305,405,779. These results indicate that the GLM Tweedie Compound Poisson approach with index parameter estimation through EQL is capable of providing flexible and representative claim reserve estimates for complex claim data characteristics.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Cadangan Klaim, Extended Quasi-likelihood, Generalized Linear Model, IBNR, Tweedie Compound Poisson, Claim Reserve
Subjects:	Q Science Q Science > QA Mathematics Q Science > QA Mathematics > QA274.2 Stochastic analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Dwiani Novita Sari
Date Deposited:	16 Jul 2026 08:30
Last Modified:	16 Jul 2026 08:30
URI:	http://repository.its.ac.id/id/eprint/135214

Actions (login required)

View Item