2.3 Define, Train, Evaluate, and Use ML Models ASSOCIATE-DATA-PRACTITIONER Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   After creating a BigQuery remote connection to Vertex AI, what does a user need to do to invoke a pretrained Google LLM from BigQuery?

   • AExport data to a CSV file and upload it to Vertex AI
   • BCreate a custom model in Vertex AI
   • CTrain a new model in BigQuery ML
   • DUse the ML.GENERATE_TEXT function in a SQL query
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A data science team is developing a model to predict customer lifetime value. They have limited historical data and need to decide between using a simple linear regression model or a more complex gradient boosted model. They also need to evaluate the model's performance before deployment. What is the best approach?

   • AUse a pretrained LLM to predict customer lifetime value.
   • BImmediately use a gradient boosted model because it will always perform better.
   • CStart with a simple linear regression model, evaluate it, and only try a more complex model if the simple one underperforms.
   • DSkip evaluation and deploy the model directly to production.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   In the ML project lifecycle, what is the primary purpose of the model evaluation phase?

   • ATo assess the model's performance using metrics like accuracy or RMSE
   • BTo generate predictions on new data
   • CTo collect and clean the training data
   • DTo deploy the model to production
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A retail company wants to predict next-month sales for each of its 500 stores. The data team is comfortable with SQL and wants to keep the entire workflow inside BigQuery, including training and generating forecasts. They do not need to export data to another tool. Which approach should they take?

   • AWrite a custom Python model in Vertex AI and schedule batch predictions to write results to BigQuery.
   • BUse a pretrained Google LLM via a BigQuery remote connection to generate sales forecasts.
   • CUse Vertex AI AutoML Tables to train a model, then export predictions to BigQuery.
   • DUse BigQuery ML with CREATE MODEL and a forecasting model type, then call ML.FORECAST.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which of the following represents the correct order of phases in a standard ML project lifecycle?

   • APrediction → Model evaluation → Data collection → Model training
   • BModel evaluation → Data collection → Model training → Prediction
   • CData collection → Model training → Model evaluation → Prediction
   • DModel training → Data collection → Prediction → Model evaluation
   Reveal Solution Discussion Select an answer first