PROPOSED NEW MODEL OF CHURN PREDICTION USING BIG DATA ANALYTIC AT PT. TELEKOMUNIKASI INDONESIA

• Main Author: Satrio Pratomo, Abiyan
• Format: Theses
• Language: Indonesia
• Subjects: Manajemen umum
• Online Access: https://digilib.itb.ac.id/gdl/view/48100
• Institution: Institut Teknologi Bandung

The digital era has penetrated various angles in our lives, all forms of IoT will accelerate the growth of data produced. Telkom as a state-owned company engaged in telecommunication industry already aware and implement big data analytic to enhance their business. Telkom as a market leader in industry has a trend of IndiHome customer churn that potentially will be increasing for each month and need to focus more in retain their customer than acquire the new one. Telkom has been implemented bigdata analytic to predict customer churn using churn prediction model. However, the churn prediction model considered out of date and cannot represent the latest characteristic of IndiHome customer. Therefore, this research aims to create a new model of churn prediction and give the insight for Telkom to retain their IndiHome customer. The variables of churn determined based on Brand Equity theory and External Analysis, that consist of Subscribing Time, Service Usage, Network Performance, Mean Time to Repair, Billing and Payments, Network Details as independent variables, and Churn as dependent variable. Total of independent variable used consist of 105 attributes. All of those attributes then processed using statistical approach such as t-test, chi-square, rfe, and logistic regression for modeling in python programming language and selected the top 8 most significant attributes that influenced customer churn. Those top 8 attributes then be used to predict customer churn and evaluated using confusion matrix. The performance of this model can be considered as good because the sensitivity is 60% to identify churn out of the actual churner customer. Telkom can get the insight that Service Usage variable is currently the variable they must be priorities on. The majority potential reason/symptom of customer churn is the substitute/competitor internet product more promising. Therefore, Telkom can be focus on following-up customer if any specific action required or to find out on what competitor’s/substitute’s internet product offer might be in order to retain the customer.