Can you explain the concept of bias-variance tradeoff in machine learning and the importance of considering it?