SUSILA, MUKTAR REDY (2016) PEMODELAN REGRESI LOGISTIK BINER BIVARIAT BAYESIAN UNTUK RESPON YANG UNBALANCE Studi Kasus: Konsumen Produk Low Price Software Antivirus Perusahaan ‘X’. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
1314201049-Abstract.pdf - Published Version Download (389kB) | Preview |
Preview |
Text
1314201049-Master Thesis.pdf - Published Version Download (5MB) | Preview |
Preview |
Text
1314201049-conclusion.pdf - Published Version Download (521kB) | Preview |
Abstract
Tujuan dari penelitian ini adalah memodelkan regresi logistik biner
Bayesian dan regresi logistik biner bivariat Bayesian pada kasus konsumen
Perusahaan ‘X’. Perusahaan ‘X’ merupakan suatu perusahaan yang beroperasi
dengan sambungan internet (Cloud Based Software). Perusahaan tersebut tengah
menyelesaikan masalah perilaku konsumen dalam perihal pembelotan produk dan
jawaban perpanjangan kontrak. Model regresi logistik biner yang memiliki dua
variabel dependen yang saling berkaitan dapat dimodelkan menjadi satu model
yang disebut regresi logistik biner bivariat. Keuntungan dari model regresi logistik
biner bivariat yaitu odds ratio yang diperoleh menggambarkan hubungan
berpasangan antara dua variabel respon biner. Metode Bayesian merupakan metode
yang menggunakan informasi-informasi sebelumnya dalam bentuk distibusi
probabilitas. Metode tersebut sering digunakan untuk pemodelan pada saat sampel
dengan variabel dependen yang unbalance. Penelitian ini menghasilkan sampel
parameter yang telah diiterasi dan belum memenuhi sifat strongly ergodic untuk
data sebanyak 500000. Sehingga dilakukan sampling dari data keseluruhan.
Didapatkan lima model regresi logistik biner Bayesian dan model regresi logistik
biner bivariat Bayesian. Model tersebut merupakan model dari sampel. Secara
univariat, variabel yang konsiten mempengaruhi Pembelotan dan Jawaban Kontrak
adalah Akumulasi Update dan Status Pengiriman e-mail. Secara bivariat, variabel
yang memberikan pengaruh terhadap kedua variabel respon adalah Status
Pengiriman. Hasil klasifikasi yang diperoleh menunjukan bahwa regresi logistik
biner univariat maupun bivariat Bayesian belum mampu menangkap efek dari
variabel Jawaban Kontrak yang unbalance.
=====================================================================================================
The aim of this study is to resolve the case in 'X' company by using Bayesian
binary logistic regression and Bayesian bivariate binary logistic regression models.
The 'X' Company is a company that operates with internet connection (Cloud-Based
Software). The company is solving customer loyalty issues in term of customer
defection and contract renewal. Binary logistic regression model that has two
interrelated dependent variables can be modeled into one model called bivariate
binary logistic regression. The advantage of the bivariate binary logistic regression
model is that the odds ratio can describes the relationship between the two pairs of
binary response variables. One of the important stages in the modeling is parameter
estimation. Common method for parameter estimation in a logistic regression
model is maximum likelihood. Bayesian method is a method that uses prior
information in the form of probability distribution. The method is often used for
modeling when the sample has unbalance dependent variable. This research yields
parameter estimates that have not strongly ergodic after iteration with 500.000
observations. Then five samples are drawn to form five Bayesian binary logistic
regression and Bayesian bivariate binary logistic regression models. Those models
are formed based on each samples. In term of univariate, variables that consistently
affect the defection and contract response are update accumulation and e-mail
delivery status. In term of bivariate, variables that affect both response variables is
the delivery status. The classification results showed that both Bayesian bivariate
and univariate binary logistic regression have not been able to capture the effects
of unbalance contract response.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | RTSt 519.536 Sus p |
| Uncontrolled Keywords: | Regresi Logistik Biner, Regresi Logistik Biner Bivariat, Bayesian, Unbalance. |
| Subjects: | Q Science > QA Mathematics Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis |
| Depositing User: | Mr. Tondo Indra Nyata |
| Date Deposited: | 10 Jan 2017 06:45 |
| Last Modified: | 27 Dec 2018 07:24 |
| URI: | http://repository.its.ac.id/id/eprint/1440 |
Actions (login required)
 |
View Item |