A linear regression model shows very high accuracy on the training set but performs poorly on new, unseen data. What action should the ML engineer take to address this problem?

The model is overfitting, meaning it has learned the training data too well, including noise, and thus generalizes poorly to new data. Cross-validation helps assess how the model will generalize by training and testing on different subsets of the data. Regularization (e.g., L1 or L2) penalizes large coefficients, simplifying the model and reducing overfitting. Amazon SageMaker Experiments is ideal for tracking these changes and comparing the performance of different model versions, ensuring the chosen solution effectively addresses the overfitting. Increasing model complexity would worsen overfitting. Deploying an overfit model is premature and will lead to poor production performance. Increasing the training dataset size can help, but without addressing the model's tendency to overfit, it might still perform poorly, especially if the model is too complex for the underlying patterns.