Developing ML Models Using BigQuery ML or AutoML on Gemini Enterprise Agent Platform PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A logistics company wants to forecast daily shipment volumes for the next 60 days using BigQuery ML. They have historical daily shipment data for the past 3 years in a table with columns `date` and `daily_shipments`. The data shows a clear weekly seasonality and some holiday effects. Which BigQuery ML approach should they use?

   • ATrain a logistic regression model with a binary label indicating whether shipments exceeded a threshold
   • BTrain an ARIMA_PLUS model with `date` as the time series column and `daily_shipments` as the target
   • CTrain a linear regression model with `date` as a feature and `daily_shipments` as the label
   • DTrain a k-means model to cluster days by shipment volume and then predict the cluster for future dates
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A media company wants to segment its users into groups for targeted content recommendations. They have a table with user features: `watch_time_minutes`, `genre_preference`, `device_type`, and `subscription_tier`. They want to use BigQuery ML k-means clustering. The team is debating whether to standardize the numeric features before training. What is the best practice?

   • ADo not standardize; k-means handles raw values well
   • BOne-hot encode all categorical features and then standardize the resulting columns
   • CStandardize numeric features using ML.STANDARD_SCALER in the TRANSFORM clause
   • DUse ML.MIN_MAX_SCALER to scale features to a fixed range
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   A company wants to use Vertex AI AutoML to train a text classification model to categorize customer support tickets into topics. They have a dataset of 10,000 labeled text samples. They are considering whether to use AutoML or a custom-trained model. The team has limited ML expertise and wants to minimize manual effort. They also need to deploy the model for online predictions. What should they do?

   • AUse a custom training job with a pre-trained BERT model and fine-tune it
   • BUse Vertex AI Model Garden to deploy a pre-trained text classification model without any training
   • CUse BigQuery ML to train a logistic regression model on the text data
   • DUse Vertex AI AutoML for text classification, which automatically handles feature engineering and model selection, then deploy the model to an endpoint
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A financial services company is building a BigQuery ML model to predict fraudulent transactions (binary). The dataset is highly imbalanced (99% non-fraud, 1% fraud). The team wants to improve recall for the fraud class without sacrificing too much precision. They have already engineered features like transaction amount, location, and time. Which additional step should they take?

   • AUse class weights in the logistic regression model to penalize misclassifications of the minority class
   • BUse ML.SMOTE to generate synthetic fraud samples
   • CUse ML.UNDERSAMPLER to randomly remove non-fraud transactions
   • DUse ML.FEATURE_IMPORTANCE to identify and remove features that are not useful
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   A retail company wants to use BigQuery ML to predict which customers are likely to churn in the next quarter. The target variable is a binary label indicating churn (yes/no). Which BigQuery ML model type should be selected?

   • ALogistic regression
   • BK-means
   • CLinear regression
   • DARIMA_PLUS
   Reveal Solution Discussion Select an answer first