You are a machine learning specialist tasked with improving a customer churn prediction model for a telecom company. The current models include gradient boosting, neural networks, and k-nearest neighbors, each with unique strengths and weaknesses in accuracy and generalization. Which ensemble method would most effectively enhance the model's accuracy and generalization by combining the diverse strengths of these individual models?