You are configuring an AutoML model and encounter an error. What could primarily cause the model to not train?
AInsufficient training data
BHigh accuracy threshold
CStrong regularization settings
DUnverified Google account
Explanation
Insufficient training data is the most common hindrance for model training, while other options are less likely causes.
Q22
Which service is best for training machine learning models at scale?
AGoogle Cloud AI Platform
BGoogle BigQuery
CGoogle Compute Engine
DGoogle Cloud Storage
Explanation
Google Cloud AI Platform is designed for scaling ML models, while the others serve different functions.
Q23
A company needs to deploy a machine learning model with minimal downtime. What should they implement?
ABlue/Green deployment strategy
BCanary release only
CRolling deployment with no monitoring
DStatic IP address for the service
Explanation
A Blue/Green deployment minimizes downtime by switching traffic smoothly, while others may introduce risks.
Q24
You are configuring AutoML for a text classification task. What happens if your dataset is highly imbalanced?
AModel performs optimally
BIncreased risk of overfitting
CBias towards the dominant class
DOnly minority class predictions improve
Explanation
Imbalanced datasets lead models to favor the dominant class, affecting overall accuracy.
Q25
Which Google Cloud service is primarily used for building machine learning models?
ACloud AutoML
BCloud Storage
CBigQuery
DCloud Functions
Explanation
Cloud AutoML is specifically designed for building ML models, while the others serve different functions.
Q26
A company needs to preprocess their data for a ML project. What would be the best service to use?
ABigQuery Data Transfer Service
BAI Platform Pipelines
CDataproc
DDataflow
Explanation
Dataflow is ideal for real-time data preprocessing, while others focus on different aspects of data management or model deployment.
Q27
You are configuring a model for predictions using the AI Platform. What happens if the selected model is larger than the maximum deployment limit?
ADeployment will succeed with warnings.
BDeployment will fail with an error.
CThe model will be downsized automatically.
DThe model will be queued for resizing.
Explanation
Model deployment fails if it exceeds the limit, while others suggest incorrect behaviors.
Q28
Which service should you use for real-time data processing?
ACloud Dataflow
BCloud Storage
CCloud Pub/Sub
DBigQuery
Explanation
Cloud Dataflow is designed for real-time data processing; the others do not provide real-time capabilities.
Q29
A company needs to build a machine learning model to predict customer churn using historical transaction data. What metric should primarily guide model evaluation?
AAccuracy
BF1 Score
CAUC-ROC
DMean Squared Error
Explanation
AUC-ROC is crucial for classification tasks like churn prediction to measure trade-offs between true and false positives; accuracy may mislead in imbalanced datasets.
Q30
You are configuring a Google Cloud AI Platform model version for deployment. What happens when you set the traffic allocation to zero?
AThe model becomes inactive
BTraffic is directed to another version
CNo predictions can be made
DThe model is deleted
Explanation
Setting traffic allocation to zero directs traffic to other active model versions, maintaining service availability; the model remains deployed.