Abstract: For any service-providing organizations, churners have always been a big issue. It increases the company's cost and lowers down the profit rate. Commonly, customer deterioration can be identified when they instigate the process of service termination. At the same time, the people and the organizations that provide the data residing on the government databases and the agencies who sponsor the collection of such details are becoming increasingly conscious that extended analytical capabilities also deliver tools that endanger the confidentiality of data records. Nevertheless, using predictive analysis using customer's previous service usage, service performance, spending, and other behavior patterns, the possibility of whether a customer wants to discontinue service can be nailed. In this paper, the writers address the problem of churn analysis, assuming a scenario in which a organization owning confidential databases wishes to run a churn analysis technique on the union of their databases without exposing any unnecessary information. The paper aims to predict whether a customer will churn shortly or not based on the predictive analysis using billing data of a telecom company.

Keywords : Random Forest, Churn , Employee-Churn

| DOI: 10.17148/IARJSET.2021.81151