Choosing a Meta-Classifier for Stacking

Imagine you're a data scientist at Google tasked with creating a spam detection system to differentiate between spam and non-spam emails based on their content. You've experimented with various algorithms such as SVM and Random Forest but haven't achieved the desired accuracy. To enhance performance, you decide to implement a stacking approach by integrating these models. Which classifier would be optimal as the meta-classifier in your stacking framework, and what is your rationale for this choice?