Prediksi Status CT-0 Menggunakan Binary dan Firth’s Logistic Regression Pada Unit Payment Collection PT Telkom Indonesia

Firdausiyah, Marshelia (2026) Prediksi Status CT-0 Menggunakan Binary dan Firth’s Logistic Regression Pada Unit Payment Collection PT Telkom Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221122-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

PT Telkom Indonesia menerapkan sistem pascabayar pada pelanggan Small Medium Enterprise (SME) yang berpotensi menimbulkan piutang tak tertagih, khususnya pada pelanggan yang berstatus CT-0 (Change Tarif 0), yaitu kondisi ketika tagihan pelanggan dinolkan akibat tunggakan selama dua bulan berturut-turut. Tingginya risiko pelanggan CT-0 yang menghilang tanpa membayar tunggakan mencerminkan fenomena silent churn yang berpotensi merugikan perusahaan. Penelitian ini bertujuan untuk mengidentifikasi faktor-faktor yang memengaruhi status CT-0 serta membangun model prediksi yang dapat digunakan sebagai dasar prioritas reminder pembayaran dan mapping CTB. Data yang digunakan adalah data C3MR pelanggan SME Witel Suramadu pada Bulan Januari 2026 dengan variabel respon status CT-0 Bulan Maret 2026, yang terdiri dari 58.620 observasi dengan proporsi CT-0 sebesar 1,8% sehingga tergolong kasus rare event. Metode yang digunakan adalah Binary Logistic Regression dengan pendekatan Maximum Likelihood Estimation (MLE) dan Firth’s Logistic Regression dengan pendekatan Penalized Maximum Likelihood Estimation (PMLE). Kedua model menghasilkan performa yang hampir identik, namun model Firth dipilih sebagai model terbaik karena lebih stabil dalam menangani data rare event dan mengatasi potensi separation. Model Firth dengan threshold optimal sebesar 0,02 dan menghasilkan akurasi 84,3%, sensitivitas 64%, spesifisitas 84,7%, dan AUC sebesar 80,7%. Model tersebut menghasilkan pemeringkatan probabilitas CT-0 tiap pelanggan yang dapat dimanfaatkan oleh unit payment collection untuk menentukan prioritas reminder pembayaran secara lebih terarah untuk menekan potensi peningkatan pelanggan yang CT-0 pada bulan berikutnya.
===================================================================================================================================
PT Telkom Indonesia applies a postpaid billing system to its Small Medium Enterprise (SME) customers, which carries the risk of uncollectible receivables, particularly among customers classified as CT-0 (Change Tarif 0), a condition in which a customer's bill is zeroed out due to two consecutive months of unpaid dues. The high risk of CT-0 customers disappearing without settling their arrears reflects a silent churn phenomenon that may result in significant financial losses for the company. This study aims to identify factors influencing CT-0 status and to develop a predictive model that can serve as the basis for prioritizing payment reminders and CTB mapping. The data used consists of C3MR records of SME customers at Witel Suramadu for January 2026, with CT-0 status in March 2026 as the response variable, comprising 58.620 observations with a CT-0 proportion of 1,8%, thereby qualifying as a rare event case. The methods applied are Binary Logistic Regression using Maximum Likelihood Estimation (MLE) and Firth's Logistic Regression using Penalized Maximum Likelihood Estimation (PMLE). Both models yielded nearly identical performance; however, the Firth model was selected as the best model due to its greater stability in handling rare event data and its ability to address potential separation issues. The Firth model with an optimal threshold of 0,02 achieved an accuracy of 84,3%, sensitivity of 64%, specificity of 84,7%, and an AUC of 80,7%. The resulting model produces a probability-based ranking of CT-0 risk for each customer, which can be utilized by the payment collection unit to determine intervention priorities more effectively and reduce the likelihood of an increase in CT-0 customers in the following month.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Binary Logistic Regression, CT-0, Firth's Logistic Regression, rare event, Small Medium Enterprise
Subjects:	H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Marshelia Firdausiyah
Date Deposited:	31 Jul 2026 08:34
Last Modified:	31 Jul 2026 08:34
URI:	http://repository.its.ac.id/id/eprint/141179

Actions (login required)

View Item