With the advancement of electronic markets platforms, customers show different purchase behaviors. Since they have a wide range of choices and low exit barriers, customer movement from one digital store to another has become a natural problem for every business. Customers are the most valuable property of every enterprise. Thus, Suppliers have to tie a strong relationship with their customers to prevent their leaving by providing the most appropriate produces and services based on their desires and trying to satisfy them. Customer churn prediction (CCP) is one of the customer relationship management (CRM) strategies to estimate the probability of their abandonment. Marketers use CCP to attract visitors, engage them in websites activities, convert them into customers, and retain them for a long time. Internet banking transactions dataset is a dependable resource to analyze customer interactions and their churn behaviors. Transactions data scrutinizes the customer’s specifications and their payment details. In this paper, we introduced a meta-classifier algorithm to predict customer churn behavior according to the transactions data. We applied the four most used supervised classification algorithms. Then, we improved their performance by hyperparameters and tuning. We also performed RFECV feature selection to extract and rank the most critical variables. The experimental results represent that combining various machine learning developed algorithms in a funnel of meta-classifier can extract the highest prediction accuracy.