No abstract

Abstract. This article discusses the classification in predicting the sustainability status of the health insurance customer policy of PT. X uses the Naïve Bayes Classifier Algorithm. In predicting the Naive Bayes Classifier Algorithm, it uses the concepts and theories of data mining in the literature related to insurance by calculating the probability of each class of variables using the Bayes theorem in describing the performance of a model or algorithm specifically using the Confusion Matrix. To be able to predict the decisions of health insurance customers in the policy sustainability status, a method of data analysis of registered insurance customers is needed. The data used is data obtained from the insurance company PT. X. The data contains customer information data in the form of 9 variables (Policy Number, Smoking Status, Gender, Age, Marital Status, Dependents, Monthly Premiums, Current Status / whether or not premium payments and insurance policy renewal status). The results of the application of the Naïve Bayes Classifier Algorithm show that the algorithm is quite good in predicting the status of the policy extension of the insured health insurance PT. X, with an average accuracy of 85.82%, an average precision of 96.10% and an average recall of 93.55. Abstrak. Artikel ini membahas tentang klasifikasi dalam memprediksi status keberlanjutan polis nasabah asuransi kesehatan PT. X menggunakan Algoritma Nave Bayes Classifier. Dalam memprediksi Algoritma Naive Bayes Classifier menggunakan konsep dan teori data mining dalam literatur yang berhubungan dengan asuransi dengan menghitung probabilitas setiap kelas variabel menggunakan teorema Bayes dalam menggambarkan kinerja suatu model atau algoritma secara khusus menggunakan Confusion Matrix . Untuk dapat memprediksi keputusan nasabah asuransi kesehatan dalam status kesinambungan polis, diperlukan suatu metode analisis data nasabah asuransi yang terdaftar. Data yang digunakan adalah data yang diperoleh dari perusahaan asuransi PT. X. Data tersebut berisi data informasi nasabah berupa 9 variabel (Nomor Polis, Status Merokok, Jenis Kelamin, Usia, Status Perkawinan, Tanggungan, Premi Bulanan, Status Lancar/tidaknya pembayaran premi dan status perpanjangan polis asuransi). Hasil penerapan Algoritma Naïve Bayes Classifier menunjukkan bahwa algoritma tersebut cukup baik dalam memprediksi status perpanjangan polis dari tertanggung asuransi kesehatan PT. X, dengan rata-rata akurasi 85,82%, presisi rata-rata 96,10% dan rata-rata recall 93,55.

This paper will discuss the modeling of claim frequency from Indonesian auto insurance using the generalized Poisson-Lindley linear model. This modeling method assumes that the data of claim frequency are from populations that follow generalized Poisson-Lindley distribution. Generalized Poisson-Lindley linear model is an alternative to modeling count data that contains overdispersion. The parameters in the generalized Poisson-Lindley linear model can be estimated using the maximum likelihood estimation method through Newton Raphson's iteration numerical method. The data are the secondary data took from XYZ Company for the 2013 policy which is overdispersed. The data contains policyholder partial loss claims for comprehensive motor vehicle insurance products. From the research conducted it was concluded that the data is suitable to be modeled with generalized Poisson-Lindley linear models and produce better models than ordinary Poisson linear modeling because of produced the smaller AIC value. Of the 3 predictor variables that are modeled on the frequency of claims, 2 variables influenced they are the use variable and vehicle brand variable.

No abstract

<p><strong>Abstract. </strong>This paper discusses the method of limited-fluctuation credibility, also known as classic credibility. Credibility theory is a technique for predicting future premium rates based on past experience data. Limited fluctuation credibility consists of two credibility, namely full credibility if Z = 1 and partial credibility if Z &lt;1. Full credibility is achieved if the amount of recent data is sufficient for prediction, whereas if the latest data is insufficient then the partial credibility approach is used. Calculations for full and partial credibility standards are used for loss measures such as frequency of claims, size of claims, aggregate losses and net premiums. The data used in this paper is secondary data recorded by the company PT. XYZ in 2014. This data contains data on the frequency of claims and the size of the policyholder's partial loss claims for motor vehicle insurance products category 4 areas 1. Based on the results of the application, the prediction of pure premiums for 2015 cannot be fully based on insurance data for 2014 because the credibility factor value is less than 1. So based on the limited-fluctuation credibility method, the prediction of pure premiums for 2015 must be based on manual values for pure premiums as well as insurance data for 2014. If manual values for pure premium is 2,000,000 rupiah, then the prediction of pure premium for 2015 is 1,849,342 rupiah.</p><p><strong>Keywords</strong><strong>: </strong>limited fluctuation credibility, full credibility, partial credibility and partial loss</p>