Data science is crucial for analytics and prediction in the telecommunication industry.
Customer churn prediction is becoming progressively important. While machine learning methods
are regularly utilized for predicting churn, their performance can be improved due to the
complexity of consumer data structures. Managers lose trust when findings are difficult to
interpret .This study utilizes data preprocessing techniques. The various elements of benchmarked
data collecting can impact interpretability since imbalanced and feature scaling issues. Therefore,
this study develops customer churn prediction model for those complexity issues. After training
the model, the operator analyzes the data to understand its performance. To maximize
interpretability, consumers are clustered based on behavioral factors. Clustering is grouping data
points with similar features to maximize similarity between members. Additionally, they share few
similarities with members of other groups. Using homogeneous group members improves
classification algorithm prediction performance. Various algorithms, including logistic regression,
support vector machine, random forest, Ada-boost, and multilayer perceptron, were tested before
and after hyperparameter adjustment to achieve optimal prediction performance.