Build and evaluate click prediction models

Build a click-through prediction model from the features above. Start with a trivial baseline classifier, then train and compare logistic regression, random forest, and gradient-boosted trees. Explain why you selected each algorithm and why you did not choose plausible alternatives (e.g., SVM, Naive Bayes, simple neural networks), discussing bias–variance, interpretability, and computational trade-offs. Describe your cross-validation strategy and the hyperparameters you would tune for each model. Choose evaluation metrics appropriate for a roughly class-balanced dataset (e.g., ROC AUC, log loss, PR AUC, accuracy) and justify why these are preferred over others; explain what would change if classes were imbalanced. Finally, outline what additional improvements you would pursue with more time.