Penerapan Model Discrete Time Logistic pada Kasus Pelanggan Telekomunikasi

Salsabila Salsabila, Abdul Kudus

Abstract


Abstract. The discrete time logistic model is an alternative solution to the survival analysis which assumes  proportionality all the time  for each covariate. Therefore, this model can be used for various conditions where the time dependent variables and time independent variables. Discrete time logistic model has easy to analyse because it uses logistic regression analysis with converting survival data from person oriented to person period. Parameters in the discrete time logistic model are estimated using the maximum likelihood method through the Fisher Scoring iteration. The application of the discrete time logistic model to the history data of telecommunications customers of Telco Companies with events where the customer is a churn or not. From this research, it can be concluded that the substantive covariates that influence the occurrence of churn are for demographic information is partners, for customer account information is contracts every one year, contracts every two years, paperless billing, payment methods by post and e-check and the services provided to customers is multiplelines, online security, online backup, device protection and tech support.

Keywords: Discrete Time Logistic, Logistic Regression, Fisher Scoring, Survival, Churn

Abstrak. Model discrete time logistic merupakan alternatif penyelesaian dari analisis survival yang mengasumsikan bahwa kovariat prediktor yang digunakannya proporsional secara konstan dari waktu ke waktu. Oleh karena itu, model ini  dapat digunakan untuk kondisi dimana kovariat prediktor yang digunakannya terdiri dari time dependent variable mapun time independent variable. Pemodelan discrete time logistic memiliki kemudahan dalam proses pengolahan datanya karena menggunakan analisis regresi logistik dengan mengperluas data survival tiap individu menjadi data person period. Parameter parameter dalam model discrete time logistic diestimasi menggunakan metode kemungkinan maksimum melalui iterasi Fisher Scoring. Penerapan model discrete time logistic dilakukan pada data riwayat pelanggan telekomunikasi Perusahaan Telco dengan event dimana pelanggan mengalami kejadian churn atau tidak. Dari penelitian ini, dapat diketahui bahwa  kovariat substantif yang memengaruhi terjadinya kejadian churn adalah untuk informasi demografi yaitu status perkawinan, untuk informasi akun pelanggan berupa kontrak setiap tahun sekali, kontrak setiap dua tahun sekali, paperless billing, metode pembayaran melalui pos dan e-check dan untuk layanan yang diberikan pada pelanggan terdiri dari multiplelines, online security, device protection dan tech support.

Kata Kunci: Discrete Time Logistic, Regresi Logistik, Fisher Scoring, Survival, Churn.


Keywords


Discrete Time Logistic, Logistic Regression, Fisher Scoring, Survival, Churn

References


Carter, D. B., & Signorino, C. S. (2010). Modeling Time Dependence in Binary Data. New York.

Efron, B. (1988). Logistic Regression, Survival Analysis and the Kapalan Meier Curve. Journal of the American Statistical Association, 414-425.

Ehlers, R. (2002). Maximum Likelihood Estimation Procedures for Categorical Data. South Africa: University of Pretoria.

Hajarisman, N. (2009). Analisis Data Kategorik. Bandung: Universitas Islam Bandung.

Han, J., Kamber, M., & Pei, J. (2012). Data Mining Concepts and Techniques. Waltham: Elsevier Inc.

Hosmer, D. W., Lemeshow, S., & May, S. (2008). Applied Survival Analysis : Regression Modeling of Time-to-Event Data. New Jersey: John Willey and Sons.

Louviere, J., Hensher, A., & Swait, D. (2000). Stated Choice Method. New York: Cambridge University Press.

Potts, W. (2004). Survival Data Mining. Retrieved from Technical White Paper in Data Miners: http://www.data-miiners.com

Singer, J. D., & Willet, J. B. (1993). It's about Time: Using Discrete-Time Survival Analysis to Study Duration and the Timing. Journal of Educational Statistics, 18, 155-195. Retrieved from http://www.jstor.org/stable/1165085




DOI: http://dx.doi.org/10.29313/.v6i2.23342

Flag Counter