You’re working on a customer churn prediction model for a telecom company. The dataset contains a mix of numeric features like ‘monthly_charges’ and ‘tenure_in_months’, and categorical features like ‘contract_type’ and ‘payment_method’. You’ve noticed that the scale of the numeric features varies significantly, and you want to ensure the model isn’t biased towards features with larger magnitudes. In Azure Machine Learning designer, which module would you use to normalize these numeric columns before training, without affecting the categorical data?

Imagine you’ve trained a classification model to detect fraudulent transactions. The initial results show a high overall accuracy, but upon reviewing the confusion matrix, you find it’s failing to catch a critical number of actual fraud cases (high false negative rate). The business cost of a missed fraud case is far higher than the cost of a false alarm. What primary metric should you now prioritize and monitor to better align the model with business objectives?

Your team has built a regression model in Azure ML to predict house prices. You need to deploy this model as a real-time inference endpoint so that a web application can get instant price predictions. However, the pre-processing pipeline for the incoming data is complex, involving imputation and feature engineering. What is the most efficient Azure ML concept to package both the trained model and its associated pre-processing steps into a single deployable unit?

During a hyperparameter tuning run for a deep learning model using Azure ML’s HyperDrive, you configure the ‘bandit’ early termination policy. You set the `slack_factor` to 0.2 and `evaluation_interval` to 2. If a run’s primary metric at interval 4 is 0.85, what would the metric need to be at interval 6 for this run to NOT be terminated early, assuming it’s the best run so far?

You are preparing a large text dataset for a natural language processing task in Azure Machine Learning. The raw text column contains customer reviews with inconsistent capitalization, punctuation, and common English stop words (like ‘the’, ‘and’, ‘is’). Which combination of modules in the designer would effectively perform all the following cleaning steps: convert to lowercase, remove punctuation, and remove stop words?

A colleague has created an AutoML experiment to classify support tickets. They set the experiment timeout to 60 minutes, the primary metric to ‘accuracy’, and enabled deep learning. The experiment finished but the resulting model’s performance is unsatisfactory. Which single change is most likely to yield a better model without a disproportionate increase in compute cost or time?

You need to track a specific, custom piece of information during the training of an Azure ML model—specifically, the version of the third-party library you used for a custom data transformation. This isn’t a standard metric like accuracy or loss. Which method should you use to log this information so it’s permanently associated with the run in Azure ML studio for auditability?

When configuring compute for an Azure ML pipeline, you have a training step that requires a powerful GPU for 2-3 hours and a subsequent data validation step that uses a simple Python script and runs for 5 minutes. To optimize cost and resource utilization, what is the recommended strategy for defining compute targets for these two steps?

After deploying a model to an AKS-based real-time endpoint, you notice the response times are highly variable and occasionally very slow. You suspect the issue is due to the endpoint scaling from zero pods when no requests have been made for a while. What deployment configuration setting in Azure ML should you adjust to mitigate this ‘cold start’ problem for a production workload with intermittent but important traffic?

You’re working on a customer churn prediction model for a telecom company. The dataset contains a mix of numeric features like ‘monthly_charges’ and ‘tenure_in_months’, and categorical features like ‘contract_type’ and ‘payment_method’. You’ve noticed that the scale of the numeric features varies significantly, and you want to ensure the model isn’t biased towards features with larger magnitudes. In Azure Machine Learning designer, which module would you use to normalize these numeric columns before training, without affecting the categorical data?0