Answer-first summary for fast verification
Answer: Define preprocessing steps with BigQuery's TRANSFORM clause when creating your model. During prediction, apply ML.EVALUATE without transforming the raw input data.
The correct approach is to use BigQuery's TRANSFORM clause during model creation to define preprocessing steps, ensuring the model is trained on preprocessed data to avoid skew. During prediction, applying ML.EVALUATE without transformations on raw data maintains consistency. Other options introduce risks of inconsistency by either manually transforming data before prediction (B), not ensuring model training on preprocessed data (C), or omitting the TRANSFORM clause in model creation (D), all of which could lead to skew in predictions.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
In your role at a large real estate firm, you're preparing to analyze 6 TB of home sales data. You plan to use SQL for data transformation and BigQuery ML to build a machine learning model, aiming to make predictions on raw, untransformed data. How should you set up your workflow to prevent skew during prediction?
A
Define preprocessing steps with BigQuery's TRANSFORM clause when creating your model. During prediction, apply ML.EVALUATE without transforming the raw input data.
B
After defining preprocessing steps with BigQuery's TRANSFORM clause during model creation, transform your raw input data using a saved query before applying ML.EVALUATE for predictions.
C
Establish preprocessing logic in a BigQuery view, use this view as your training data for model creation, and apply ML.EVALUATE without transformations on raw data during prediction.
D
Process all data through Dataflow for preprocessing. During prediction, use ML.EVALUATE without additional transformations on the input data.
No comments yet.