The paper provides a comparative study of some machine learning algorithms used for customer churn prediction in terms of their effectiveness before and after hyperparameter tuning. In the models, the considered models include logistic regression, decision tree, XGBoost, Gradient Boosting Classifier and Random Forest, with accuracy as the primary performing metric. Hyperparameter tuning improved the accuracy for the greater part of the models—the Gradient Boosting Classifier showed a dramatic increase from 86.28 to 91.56%. The decision tree model improved significantly, with an accuracy now at 85.75% compared to the original value of 77.55%, while logistic regression and Random Forest only marginally improved, while XGBoost had a small drop in accuracy after optimizing its hyperparameters from 81.2 to 76.6%. These results show how optimization of hyperparameters can strengthen the predictive strength of algorithms particularly for the case of churn prediction. This research informs selection and optimal choice of models for achieving the best possible results in relation to retention strategies for customers.