Rapid technology growth has affected corporate practices. With more items and services to select from, client churning has become a big challenge and threat to all firms. We offer a machine learning-based churn prediction model for a B2B subscription-based service provider. Our research aims to improve churn prediction. We employed machine learning to iteratively create and evaluate the resulting model using accuracy, precision, recall, and F1- score. The data comes from a financial administration subscription service. Since the given dataset is mostly non-churners, we analyzed SMOTE, SMOTEENN, and Random under Sampler to balance it. Our study shows that machine learning can anticipate client attrition. Ensemble learners perform better than single base learners, and a balanced training dataset should increase classifier performance