Prediction of Churn in Telecom Service: Exploring Call Behaviors and using Machine Learning
Jae Won Choi
Jae Won Choi, College of software, Chungang University, Heugseoglo 84, Dongjaggu, Seoul, Korea.
Manuscript received on November 18, 2019. | Revised Manuscript received on 27 November, 2019. | Manuscript published on December 10, 2019. | PP: 3831-3834 | Volume-9 Issue-2, December 2019. | Retrieval Number: B7189129219/2019©BEIESP | DOI: 10.35940/ijitee.B7189.129219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Churn has a significant impact on mobile network operators and telecommunications service providers. Many studies on churn have been reported, but no one can say that they can create universal human tools for predicting churn or that we can see all the reasons for it. The purpose of this study is to derive the call behavior factors of churning customers and to find ways to reduce the churn of target customers who exhibit these call behaviors. For this, this study uses decision tree and machine learning for the prediction of churn in telecom service. Based on the analysis results, first, the fact that the total number of customers who have more than 316.7 in churn shows that the higher the number of calls, the higher the chance of churn. Second, among customers with total day minutes above 316.7, those with customer service calls above 8.5 show a high likelihood of churn among complaining customers. The overall accuracy is 91.4%. Among the customers who predicted not to be churned, the accuracy that would not be churned was 92.87%, and the accuracy that was churned was 78.4% among the customers predicted to be churned.
Keywords: Telecom service, Churn, Decision Tree, Machine learning.
Scope of the Article: Machine learning