Prediksi Time-to-Churn Pelanggan Pada Industri Telekomunikasi Menggunakan Pendekatan Analisis Survival dan Machine Learning

Gustha, Annisa Nathania Eka (2026) Prediksi Time-to-Churn Pelanggan Pada Industri Telekomunikasi Menggunakan Pendekatan Analisis Survival dan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241172-Master_Thesis.pdf Download (4MB)

Abstract

Industri telekomunikasi di Indonesia tengah menghadapi persaingan yang semakin intens seiring meningkatnya penetrasi internet dan kecenderungan pasar menuju kondisi jenuh. Dalam situasi tersebut, kemampuan perusahaan untuk mempertahankan pelanggan menjadi semakin krusial, sehingga churn pelanggan perlu dipahami tidak hanya sebagai kejadian biner, tetapi sebagai proses yang dipengaruhi oleh dimensi waktu. Penelitian ini bertujuan untuk menganalisis perilaku churn pelanggan berbasis time-to-event serta membandingkan kinerja regresi Cox Proportional Hazards dan Survival XGBoost dalam memprediksi risiko churn. Penelitian ini menggunakan dataset publik IBM Telco Customer Churn. Hasil analisis deskriptif menunjukkan tingkat churn sebesar 26,6% yang terkonsentrasi pada periode awal berlangganan, dengan median time-to-churn pelanggan churn hanya 10 bulan dibandingkan 38 bulan pada pelanggan aktif. Model Extended Cox Time-Dependent dengan seleksi variabel melalui filtering dan backward elimination menghasilkan delapan variabel prediktor signifikan yang diinteraksikan dengan fungsi waktu ln⁡(t) dan Heavyside dengan Concordance Index sebesar 0,964. Pendekatan sekuensial diterapkan dengan menggunakan hasil seleksi variabel Extended Cox sebagai input Survival XGBoost, yang menghasilkan C-Index rata-rata sebesar 0,8595. Hasil penelitian menunjukkan bahwa pendekatan survival mampu memberikan pemahaman yang lebih komprehensif terhadap dinamika risiko churn dan dapat menjadi dasar dalam perumusan strategi retensi pelanggan yang lebih tepat sasaran.
================================================================================================================================
The telecommunications industry in Indonesia is facing increasingly in-tense competition as internet penetration increases and the market trends toward sat-uration. In this situation, a company's ability to retain customers is becoming in-creasingly crucial, so customer churn needs to be understood not only as a binary event but as a process influenced by the dimension of time. This study aims to ana-lyze customer churn behavior based on time-to-event and compare the performance of Cox Proportional Hazards regression and Survival XGBoost in predicting churn risk. This study uses the IBM Telco Customer Churn public dataset. Descriptive analysis results show a churn rate of 26.6% concentrated in the early stages of the subscription period, with a median time-to-churn of only 10 months for churned customers compared to 38 months for active customers. The Extended Cox Time-Dependent model, with variable selection via filtering and backward elimination, yielded eight significant predictor variables and had a Concordance Index of 0.964. A sequential approach was applied by using the variable selection results from the Extended Cox model as input for Survival XGBoost, which produced an average C-Index of 0.8595. The results of the study show that the survival approach is able to provide a more comprehensive understanding of the dynamics of churn risk and can be the basis for formulating more targeted customer retention strategies.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	customer churn, analisis survival, Cox Proportional Hazards, Survival XGBoost, industri telekomunikasi, survival analysis, telecommunication industry
Subjects:	H Social Sciences > HA Statistics H Social Sciences > HD Industries. Land use. Labor > HD30.27 Business forecasting Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101 Telecommunication
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Annisa Nathania Eka Gustha
Date Deposited:	29 Jul 2026 03:27
Last Modified:	29 Jul 2026 03:27
URI:	http://repository.its.ac.id/id/eprint/139257

Actions (login required)

View Item