Exploring and Preprocessing Data for ML PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   An engineer must preprocess a multi-terabyte dataset stored in Google Cloud Storage. The preprocessing involves complex transformations that require custom code and can be parallelized. Which tool is most appropriate for this task?

   • AGoogle Sheets
   • BBigQuery
   • CIn-memory Python (e.g., pandas)
   • DDataflow
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   An e-commerce company has separate teams for search ranking and recommendation. Both teams use a 'user average order value' feature in their models, but they compute it differently, leading to inconsistent predictions. The ML platform team wants to consolidate this feature so both teams use the same definition and values. What should they do?

   • ALet each team keep their own computation but add a post-processing step to align the values
   • BHave the search team's pipeline write the feature to a shared BigQuery table for the recommendation team to read
   • CDocument the correct computation in a shared wiki and ask both teams to update their code
   • DCreate a single feature definition in Gemini Enterprise Agent Platform Feature Store and migrate both teams to use it
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   What is the primary purpose of a feature store in a machine learning platform?

   • ATo manage the versioning of datasets used for training and evaluation.
   • BTo execute data preprocessing pipelines on a schedule.
   • CTo provide a centralized repository for reusable, consistent features across teams and projects.
   • DTo store trained models and their metadata for deployment.
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A logistics company has a dataset with 10 million rows of shipment records. The data includes numeric fields (weight, distance), categorical fields (shipment type, region), and a free-text field with driver notes. The ML team wants to build a model to predict delivery delay. What is the most appropriate first step to organize this data for experimentation?

   • ALoad the data into a pandas DataFrame and use matplotlib to visualize all features
   • BProfile the data using BigQuery to understand distributions, missing values, and cardinality of categorical features
   • CImmediately train a baseline model with all features to see which ones are important
   • DSeparate the data into three datasets: numeric, categorical, and text, and build separate models for each
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   During data exploration for a new ML project, an engineer needs to quickly assess the distribution of a numerical column in a large tabular dataset. Which action is most appropriate for this task?

   • ALoad the entire dataset into a pandas DataFrame and use the describe() method.
   • BCreate a scatter plot of the column against the index to visualize the distribution.
   • CSample a few rows manually and estimate the distribution from the sample.
   • DUse a SQL query to compute summary statistics such as min, max, mean, and percentiles directly on the data source.
   Reveal Solution Discussion Select an answer first