Published Paper Details:

The retention of customers has become an important objective in the telecommunications industry since high churn rate damages revenue and competition. The ability to forecast who will leave enables the companies to take measures in time to retain them. This study presents an entire machine-learning method to forecast and model churn within a telecom environment. With SQL server, we collected historical information about customers (such as their personal information, service usage, billing, and previous churn). Data cleaning and transformation were undertaken with attention and critical attributes were generated. The predictive tool created was based on a Random Forest model in Python which was capable of identifying probable churners with a high level of accuracy. In conjunction with this, interactive power bi dashboards were developed to display churn trends, customer segments and trends of behavior, thus making decisions informed by data. The model was tested and found to identify key drivers of churn such as the nature of contract, monthly fees and frequency of services usage by customers. It contributed to the increase of customer retention by approximately 5% in three months.