Why is tree pruning useful in decision tree induction? Discuss how it helps improve the model's performance and prevent overfitting. What are the common methods of tree pruning, and how do they work? Additionally, what is the drawback of using a separate set of tuples to evaluate pruning? Consider the impact on the size of the training data and the overall model performance, and suggest alternative strategies that might mitigate this issue. How can these strategies help maintain the robustness and accuracy of the decision tree model?