Explain classification lifecycle and CTR modeling

You are in an interview for a Machine Learning Engineer position. Address the following machine-learning topics in an organized manner:

1. Outline a concrete bag-of-words text feature pipeline. Cover tokenization, building the vocabulary, dealing with rare or out-of-vocabulary words, sparse representation, and weighting options like raw counts or TF-IDF.
2. Explain out-of-bag (OOB) evaluation for ensemble approaches such as bagging or random forests. How do OOB samples arise, and how can they serve as a validation set?
3. Imagine you must create a binary classifier for click-through rate (CTR) prediction. Walk through the entire workflow from defining the problem to deployment, including data gathering, feature engineering, model choice, training, calibration, and evaluation.
4. More broadly, if asked to construct a classification model from the ground up, go through each major step and note suitable techniques or model options at every stage.
5. If the model's live performance degrades after deployment, how would you diagnose and troubleshoot the problem? Address causes at the model, data, serving, and product levels.

Overview: This question assesses skill in end-to-end supervised classification and production ML systems, spanning bag-of-words text feature engineering, out-of-bag ensemble evaluation, CTR prediction workflows with calibration and evaluation, and debugging deployed models in operation.