Emmanuel Chai Nzala, Kennedy Khadullo and Kevin Tole.

ABSTRACT

In this paper we address the Customer Churn Prediction Problem (CCPP). CCPP involves predicting and identifying potential churners based on historical data so that appropriate measures, interventions, or strategies can be implemented to retain or mitigate the negative impact of their departure. Our objective is to identify and scrutinize methodologies that augment the accuracy and efficacy of customer churn predictions. Through our analysis, it was revealed that automatic churn prediction encounters a challenge stemming from the inherent disparities within the dataset. Specifically, there exists a notable disproportion between the majority and minority classes, potentially leading to model bias that favors the dominant class. This synthesized literature will highlight gaps and limitations in existing research. In conclusion, this literature review represents a significant contribution towards illuminating the need for imbalance correction, thereby fostering an enhanced accuracy and facilitating the progression of future studies.

Key words: Customer Churn Prediction, Imbalance Class, Insurance, Machine Learning models.