Computer File

Pemodelan tingkat pemulihan asuransi grup cacat jangka panjang menggunakan Gradient Boosting Machine (GBM)

An insurance company needs to know the recovery rate of each client’s disability therefore, it is
necessary to build a model that can predict the rate of recovery of disability. The Society of
Actuaries (SOA) builds a predictive model for the recovery rate of disability using Decision Tree.
Decision Tree is a machine learning method to build predictive models from data. The model is
obtained by partitioning the predictor space (the space formed by independent variables) into a
number of simple regions. The partition can be represented graphically as a decision tree. In
order to obtain a more accurate prediction model than the Decision Tree method, an ensemble
method was developed. It builds a number of predictive models and then integrates these models
to get a model with better prediction performance. One popular ensemble method is Boosting.
The two popular Boosting algorithms are Gradient Boosting Machine (GBM) and AdaBoost.
This final project will discuss the modelling of long-term disability recovery rates using GBM
method. GBM is able to improve the performance of predictions resulting from Decision Tree
discussed in journal Predicting Group Long Term Disability Recovery and Mortality Rates Using
Tree Models. Mean Square Error (MSE) will be used to validate the prediction models obtained
using GBM. Then the MSE from the GBM and Decision Tree model will be compared. Based
on the simulation results it turns out that the MSE value of GBM model is smaller than the
MSE of Decision Tree model. It can be concluded that GBM can improve the performance of
the model in predicting recovery rate of disability. GBM also can be used for specific application
needs by adjusting the loss-function.
Keywords: Recovery rate, Decision Tree, Ensemble Method, Boosting, Gradient Boosting
Machine (GBM), Mean Square Error (MSE), Loss-function.

Ketersediaan

Informasi Detail

Judul Seri

No. Panggil

MAT MER p/20

Penerbit

Bandung : Program Studi Matematika Fakultas Teknologi Informasi dan Sains - UNPAR., 2020

Deskripsi Fisik

xxii, 37 p. : il. ; 30 cm

Bahasa

Indonesia

ISBN/ISSN

-

Klasifikasi

-

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

-

Info Detail Spesifik

1685 - FTIS

Pernyataan Tanggungjawab

Agnes G. Mercyana S.

Versi lain/terkait

Lampiran Berkas

Komentar