Amazon
Avaya
Cisco
CompTIA
CrowdStrike
EC-Council
EXIN
Fortinet
Google
ISC
ISC2
ITIL
Juniper
Microsoft
NVIDIA
Okta
Palo Alto Networks
Salesforce
SAP
ServiceNow
All Vendors

Google Professional Machine Learning Engineer Google Professional Machine Learning Engineer Dumps in PDF

Free Google Google Professional Machine Learning Engineer Real Questions (page: 36)

Google Professional Machine Learning Engineer PDF — 339 Questions

Your company manages a video sharing website where users can watch and upload videos. You need to create an ML model to predict which newly uploaded videos will be the most popular so that those videos can be prioritized on your company's website. Which result should you use to determine whether the model is successful?

  1. The model predicts videos as popular if the user who uploads them has over 10,000 likes.
  2. The model predicts 97.5% of the most popular clickbait videos measured by number of clicks.
  3. The model predicts 95% of the most popular videos measured by watch time within 30 days of being uploaded.
  4. The Pearson correlation coefficient between the log-transformed number of views after 7 days and 30 days after publication is equal to 0.

Answer(s): C

Explanation:

Option C is correct because it aligns evaluation with business objective: predicting the proportion of videos that become popular as measured by watch time within 30 days directly reflects user engagement and long-term value, which is a typical metric for popularity prediction in video platforms.
A) Incorrect — using uploader likes is a proxy signal and circular bias; it does not measure actual viewer engagement or video performance after upload.
B) Incorrect — 97.5% of clickbait videos by number of clicks focuses on short-term clicks, not sustained watch time, and can incentivize low-quality content.
D) Incorrect — a Pearson correlation of 0 indicates no linear relationship, not a meaningful measure of predictive success.



You are working on a Neural Network-based project. The dataset provided to you has columns with different ranges. While preparing the data for model training, you discover that gradient optimization is having difficulty moving weights to a good solution. What should you do?

  1. Use feature construction to combine the strongest features.
  2. Use the representation transformation (normalization) technique.
  3. Improve the data cleaning step by removing features with missing values.
  4. Change the partitioning step to reduce the dimension of the test set and have a larger training set.

Answer(s): B

Explanation:

Option B is correct because normalization (representation transformation) scales features to a similar range, which stabilizes and accelerates gradient-based optimization in neural networks, helping weights converge more effectively.
A) Incorrect — feature construction may help if features are weak, but it does not directly address gradient optimization stability or scaling issues.
C) Incorrect — removing features with missing values addresses data quality, not gradient optimization or feature scale; imputation or robust modeling would be more relevant.
D) Incorrect — partitioning to alter train/test sizes does not fix feature scaling or optimization dynamics and can harm evaluation validity.



Your data science team needs to rapidly experiment with various features, model architectures, and hyperparameters. They need to track the accuracy metrics for various experiments and use an API to query the metrics over time. What should they use to track and report their experiments while minimizing manual effort?

  1. Use Kubeflow Pipelines to execute the experiments. Export the metrics file, and query the results using the Kubeflow Pipelines API.
  2. Use AI Platform Training to execute the experiments. Write the accuracy metrics to BigQuery, and query the results using the BigQuery API.
  3. Use AI Platform Training to execute the experiments. Write the accuracy metrics to Cloud Monitoring, and query the results using the Monitoring API.
  4. Use AI Platform Notebooks to execute the experiments. Collect the results in a shared Google Sheets file, and query the results using the Google Sheets API.

Answer(s): A

Explanation:

Option A is correct because Kubeflow Pipelines provides a robust framework to orchestrate ML experiments and exposes an API to access run-level metrics and lineage, enabling rapid experimentation with minimal manual effort.
B is incorrect because while AI Platform Training can run experiments, exporting metrics to BigQuery and querying via the BigQuery API adds unnecessary complexity for time-series experiment metrics tracking and does not leverage a purpose-built ML experiment-tracking workflow.
C is incorrect because Cloud Monitoring is designed for infrastructure and service metrics, not experiment-level ML metrics, and using its API is not the optimal path for tracking model accuracy across experiments.
D is incorrect because AI Platform Notebooks with Google Sheets introduces manual data collection and lacks scalable, programmatic experiment tracking and versioning that a dedicated ML pipeline provides.



You work for a bank and are building a random forest model for fraud detection. You have a dataset that includes transactions, of which 1% are identified as fraudulent. Which data transformation strategy would likely improve the performance of your classifier?

  1. Write your data in TFRecords.
  2. Z-normalize all the numeric features.
  3. Oversample the fraudulent transaction 10 times.
  4. Use one-hot encoding on all categorical features.

Answer(s): C

Explanation:

Option C is correct because oversampling the minority class (fraud) addresses severe class imbalance, improving the model’s ability to learn decision boundaries for fraud cases and typically boosting metrics like recall/precision in imbalanced fraud detection with tree-based models.
A) Incorrect — Writing data in TFRecords is an I/O/performance optimization, not a predictive feature transformation that directly improves classifier performance.
B) Incorrect — Z-normalization is not always beneficial for tree ensembles like random forests, which are tree-based and less sensitive to feature scaling.
D) Incorrect — One-hot encoding can increase sparsity and dimensionality without addressing class imbalance, potentially harming model efficiency and performance.


Reference:

https://towardsdatascience.com/how-to-build-a-machine-learning-model-to-identify-credit-card- fraud-in-5-stepsa-hands-on-modeling-5140b3bd19f1



You are using transfer learning to train an image classifier based on a pre-trained EfficientNet model. Your training dataset has 20,000 images. You plan to retrain the model once per day. You need to minimize the cost of infrastructure. What platform components and configuration environment should you use?

  1. A Deep Learning VM with 4 V100 GPUs and local storage.
  2. A Deep Learning VM with 4 V100 GPUs and Cloud Storage.
  3. A Google Kubernetes Engine cluster with a V100 GPU Node Pool and an NFS Server
  4. An AI Platform Training job using a custom scale tier with 4 V100 GPUs and Cloud Storage

Answer(s): D

Explanation:

Option D is correct because AI Platform Training with a custom scale tier allows on-demand, scalable training jobs (4 V100 GPUs) using Cloud Storage for data and artifacts, minimizing idle infrastructure and cost while supporting daily retraining.
A) Incorrect — 4 V100 GPUs with local storage incurs higher cost and doesn’t leverage scalable, managed training; local storage isn’t optimal for distributed training or long-term cost efficiency.
B) Incorrect — 4 V100 GPUs with Cloud Storage still uses a fixed VM setup; no managed scaling or orchestration, increasing management overhead and cost for daily retraining.
C) Incorrect — GKE with an NFS server adds complexity and persistent storage management; not optimized for cost-efficient, single-job daily retraining.



While conducting an exploratory analysis of a dataset, you discover that categorical feature A has substantial predictive power, but it is sometimes missing. What should you do?

  1. Drop feature A if more than 15% of values are missing. Otherwise, use feature A as-is.
  2. Compute the mode of feature A and then use it to replace the missing values in feature A.
  3. Replace the missing values with the values of the feature with the highest Pearson correlation with feature
  5. Add an additional class to categorical feature A for missing values. Create a new binary feature that indicates whether feature A is missing.

Answer(s): D

Explanation:

Option D is correct because treating missingness as a signal by adding an indicator for missing values and keeping A's original category leverages the informative nature of missingness and preserves information for models that can handle categorical features.
A) Incorrect — arbitrarily dropping based on a threshold ignores the potential predictive value of A and the missingness signal.
B) Incorrect — imputing with mode can bias category distributions and ignore missingness as information.
C) Incorrect — imputing from the single most correlated feature risks leakage and ignores the explicit missingness pattern.



You work for a large retailer and have been asked to segment your customers by their purchasing habits. The purchase history of all customers has been uploaded to BigQuery. You suspect that there may be several distinct customer segments, however you are unsure of how many, and you don't yet understand the commonalities in their behavior. You want to find the most efficient solution. What should you do?

  1. Create a k-means clustering model using BigQuery ML. Allow BigQuery to automatically optimize the number of clusters.
  2. Create a new dataset in Dataprep that references your BigQuery table. Use Dataprep to identify similarities within each column.
  3. Use the Data Labeling Service to label each customer record in BigQuery. Train a model on your labeled data using AutoML Tables. Review the evaluation metrics to understand whether there is an underlying pattern in the data.
  4. Get a list of the customer segments from your company's Marketing team. Use the Data Labeling Service to label each customer record in BigQuery according to the list. Analyze the distribution of labels in your dataset using Data Studio.

Answer(s): A

Explanation:

Option A is correct because k-means clustering with BigQuery ML can perform unsupervised segmentation directly on the existing BigQuery data, and BigQuery ML can automatically determine a reasonable number of clusters (e.g., via elbow method or in-model initialization), making it efficient for discovering customer segments without labeled data.
B is incorrect because Dataprep is for data cleaning/prep, not for discovering latent segments or clustering similarities across columns; it doesn’t perform unsupervised segmentation.
C is incorrect because Data Labeling Service and AutoML Tables are for supervised learning with labeled data; without labeled segments, this isn’t appropriate for discovering inherent purchase-based segments.
D is incorrect because labeling to a fixed list from Marketing introduces predefined categories; it does not discover natural segments and adds labeling overhead.



You recently designed and built a custom neural network that uses critical dependencies specific to your organization's framework. You need to train the model using a managed training service on Google Cloud. However, the ML framework and related dependencies are not supported by AI Platform Training. Also, both your model and your data are too large to fit in memory on a single machine. Your ML framework of choice uses the scheduler, workers, and servers distribution structure. What should you do?

  1. Use a built-in model available on AI Platform Training.
  2. Build your custom container to run jobs on AI Platform Training.
  3. Build your custom containers to run distributed training jobs on AI Platform Training.
  4. Reconfigure your code to a ML framework with dependencies that are supported by AI Platform Training.

Answer(s): C

Explanation:

Option C is correct because AI Platform Training supports custom containers for distributed training, enabling large-scale training with the framework’s scheduler, workers, and servers architecture when native support is lacking. The scenario requires distributed training across multiple machines, which is accommodated by custom containers on AI Platform Training. A) is incorrect because built-in models don’t address the need for your custom, large-scale, framework-based training. B) insufficient because single-container jobs don’t cover distributed training across multiple workers/servers. D) unnecessary since you can use distributed custom containers without changing to a different framework.



PDF
Viewing page 36 of 44

Share your comments for Google Google Professional Machine Learning Engineer exam with other users:

A
AI Tutor Explanation
4/29/2026 5:23:14 AM

Question 5:

  • Correct answer: A. User4 and User1 only

  • Why this is correct:
- The Automatic Enrollment setting in Intune has MDM user scope: GroupA. Only users in GroupA can enroll devices via MDM auto-enrollment. - Device6 will be enrolled via Windows Autopilot and Intune, so enrollment is allowed only for users in GroupA. - Based on the group memberships in the scenario, User4 and User1 are in GroupA, while User2 and User3 are not. Therefore only User4 and User1 can enroll Device6.
  • Quick tip for the exam:
- Remember: MDM user scope determines who can auto-enroll devices; MAM scope controls app protection enrollment. When a new Autopilot device is enrolled, the signing-in user must be in the MDM scope.

Durban, South Africa
A
AI Tutor Explanation
4/29/2026 5:17:10 AM

Why this is correct

  • Correct answer: C. Extract the hardware ID information of each computer to a CSV file and upload the file from the Microsoft Intune admin center.

  • Why this is correct:
- Windows Autopilot requires devices to be registered by their hardware IDs (hash) before Autopilot can deploy Windows 10 Enterprise. - Collect the hardware IDs from the new Phoenix machines, save them in a CSV, and upload that CSV in the Intune/Windows Autopilot area. This maps each device to an Autopilot deployment profile. - After registration, you can assign Autopilot profiles (Windows 10 Enterprise, etc.). Other options (serial number CSV, generalizing, or Mobility settings) are not the initial Autopilot registration steps.

Durban, South Africa
A
AI Tutor Explanation
4/25/2026 1:53:46 PM

Question 7:

  • Correct answer: B — A risk score is computed based on the number of remediations needed compared to the industry peer average.

Explanation:
  • Risk360 uses a remediation-based score. It benchmarks how many actions are required to fix issues against peers, giving a relative risk posture.
  • Why not the others:
- A: Not just total risk events by location. - C: Time to mitigate isn’t the primary scoring method. - D: Not a four-stage breach scoring approach.
Note: The page text shows a mismatch (it lists D as the answer), but the study guide describes the remediation-based scoring (B) as the correct concept.

New Delhi, India
A
AI Tutor Explanation
4/25/2026 1:42:20 PM

Question 104:

  • Correct answer: D) Multi-Terabyte (TB) Range

  • Brief explanation:
- clustering keys organize data into micro-partitions to improve pruning when queries filter on those columns. - The performance benefit is most significant for very large tables; for small tables the overhead of maintaining clustering outweighs gains. - Therefore, as a best practice, define clustering keys on tables at the TB scale.

Thane, India
C
Community Helper
4/25/2026 2:03:10 AM

Q23: Fabric Admin is correct. Because Domain admin cannot create domains. Only Fabric Admin can among the given options. Q51: Wrapping @pipeline.parameter.param1 inside {} will return a string. But question requires the expression to return Int, so correct answer should be @pipeline.parameter.param1 (no {})

New Delhi, India
A
AI Tutor Explanation
4/23/2026 3:07:03 PM

Question 62:

  • Correct answer: D (per the page)

  • Note: The explanation text on the page describes option B (use ZDX score and Analyze Score to trigger the Y Engine analysis), indicating a mismatch between the stated answer and the rationale.

  • Key concept: For fast root-cause analysis, leverage telemetry and auto-correlated insights:
- Use the user’s ZDX score for AWS and run Analyze Score to activate the Y Engine, which correlates metrics across network, client, and application to pinpoint the issue quickly.
  • Why the other options are less effective:
- A: Only checks for outages; doesn’t provide actionable root-cause analysis. - C: Deep Trace helps visibility but is manual and time-consuming. - D: Packet capture is invasive and slow; not the quickest path to root cause.

Coimbatore, India
A
AI Tutor Explanation
4/23/2026 12:26:21 PM

Question 32:

  • Answer: A (2.4GHz)

  • Why: Lower-frequency signals have longer wavelengths and experience less attenuation when passing through walls and obstacles. Higher frequencies (5GHz, 6GHz) are more easily blocked by walls. NFC operates over very short distances and is not meant to penetrate walls. So 2.4 GHz best penetrates physical objects like walls.

Allen, United States
A
AI Tutor Explanation
4/21/2026 8:48:36 AM

Question 3:

  • False is the correct answer (Option B).

Why:
  • In Snowflake, a database is a metadata object that exists within a single Snowflake account. Accounts are isolated—there isn’t one database that lives in multiple accounts.
  • You can access data across accounts via data sharing or database replication, but these create separate database objects in the other accounts (e.g., a database in the consumer account created from a share), not a single shared database across accounts.

So a single database cannot exist in more than one Snowflake account.

Thane, India
A
Anonymous User
4/16/2026 10:54:18 AM

Question 1:

  • Correct answer: E — date = sys.argv[1]
  • Why this is correct:
- When a Databricks Job passes parameters to a notebook, those parameters are supplied to the notebook's Python process as command-line arguments. The first argument after the script name is sys.argv[1], so date = sys.argv[1] captures the passed date value directly.
  • How it compares to other options:
- date = spark.conf.get("date") reads from Spark config, not from job parameters. - input() waits for user input at runtime, which isn’t how job parameters are provided. - date = dbutils.notebooks.getParam("date") would work if the notebook were invoked via dbutils.notebook.run with parameters, not

Innisfil, Canada
A
Anonymous User
4/15/2026 4:42:07 AM

Question 528:

  • Correct answer: NSG flow logs for NSG1 (Option B)

  • Why:
- Traffic Analytics uses NSG flow logs to analyze traffic patterns. You must have NSG flow logs enabled for the NSGs you want to monitor. - An Azure Log Analytics workspace is also required to store and query the traffic data. - Network Watcher must be available in the subscription for traffic analytics to function.
  • What to configure (brief steps):
- Ensure Network Watcher is enabled in the East US region (for the subscription/region). - Enable NSG flow logs on NSG1. - Ensure a Log Analytics workspace exists and is accessible (read/write) so Traffic Analytics can store and query logs.
  • Why other options aren’t correct:
- “Diagnostic settings for VM1” or “Diagnostic settings for NSG1” alone don’t guarantee flow logs are captured and sent to Log Analytics, which Traffic Analytics relies on. - “Insights for VM1” is not how Traffic Analytics collects traffic data.

Hamburg, Germany
A
Anonymous User
4/15/2026 2:43:53 AM

Question 23:
The correct answer is Domain admin (option B), not Fabric admin.

  • Domain admin provides domain-level management: create domains/subdomains and assign workspaces within those domains, which matches the tasks while following least privilege.
  • Fabric admin is global-level access and is more privileges than needed for this scenario (it would grant broader control across the Fabric environment).

Kottayam, India
A
Anonymous User
4/14/2026 12:31:34 PM

Question 2:
For question 2, the key concept is the Longest Prefix Match. Routers pick the route whose subnet mask is the most specific (largest prefix length) that still matches the destination IP.
From the options:

  • A) 10.10.10.0/28 ? 10.10.10.0–10.10.10.15
  • B) 10.10.13.0/25 ? 10.10.13.0–10.10.13.127
  • C) 10.10.13.144/28 ? 10.10.13.144–10.10.13.159
  • D) 10.10.13.208/29 ? 10.10.13.208–10.10.13.215

The destination Host A’s IP must fall within 10.10.13.208–10.10.13.215 for the /29 to be the best match. Since /29 is the longest prefix among the matching options, Router1 will use 10.10.13.208/29.
Thus, the correct answer is D.

Canada
S
srameh
4/14/2026 10:09:29 AM

Question 3:

  • Correct answer: Phase 4, Post Accreditation

  • Explanation:
- In DITSCAP, the four phases are: - Phase 1: Definition (concept and requirements) - Phase 2: Verification (design and testing) - Phase 3: Validation (fielding and evaluation) - Phase 4: Post Accreditation (ongoing operations and lifecycle management) - The description—continuing operation of an accredited IT system and addressing changing threats throughout its life cycle—fits the Post Accreditation phase, which covers operations, maintenance, monitoring, and reauthorization as threats and environment evolve.

France
O
onibokun10
4/13/2026 7:50:14 PM

Question 129:
Correct answer: CNAME

  • A CNAME record creates an alias for a domain, so newapplication.comptia.org will resolve to whatever IP address www.comptia.org resolves to. This ensures both names point to the same resource without duplicating the IP.
  • Why not the others:
- SOA defines authoritative information for a zone. - MX specifies mail exchange servers. - NS designates name servers for a zone.
  • Notes: The alias name (newapplication.comptia.org) should not have other records if you use a CNAME for it, and CNAMEs aren’t used for the zone apex (root) domain. This scenario uses a subdomain, so a CNAME is appropriate.

Swindon, United Kingdom
A
Anonymous User
4/13/2026 6:29:58 PM

Question 1:

  • Correct answer: C

  • Why this is best:
- Uses OS Login with IAM, so SSH access is granted via Google accounts rather than distributing per-user SSH keys. - Granting the compute.osAdminLogin role to a Google group gives admin access to all team members in a centralized, auditable way. - Access is auditable: Cloud Audit Logs show who accessed which VM, satisfying the security requirement to determine who accessed a given instance.
  • How it works:
- Enable OS Login on the project/instances (enable-oslogin metadata). - Add the team’s

Germany
A
Anonymous User
4/13/2026 1:00:51 PM

Question 2:

  • Answer: D. Azure Advisor

  • Why: To view security-related recommendations for resources in the Compute and Apps area (including App Service Web Apps and Functions), you use Azure Advisor. Advisor surfaces personalized best-practice recommendations across resources, including security, and shows which resources are affected and the severity.

  • Why not the others:
- Azure Log Analytics is for ad-hoc querying of telemetry, not for viewing security recommendations. - Azure Event Hubs is for streaming telemetry data, not for security recommendations.
  • Quick tip: In the portal, navigate to Azure Advisor and check the Security recommendations for App Services to see actionable items and affe

    Brazil
D
Don
4/11/2026 5:36:42 AM

Recommend using AI for Solutions rather the Answer(s) submitted here

Hamburg, Germany
M
Mogae Malapela
4/8/2026 6:37:56 AM

This is very interesting

Gaborone, Botswana
A
Anon
4/6/2026 5:22:54 PM

Are these the same questions you have to pay for in ExamTopics?

Amsterdam, The Netherlands
L
LRK
3/22/2026 2:38:08 PM

For Question 7 - while the answer description indicates the correct answer, the option no. mentioned is incorrect. Nice and Comprehensive. Thankyou

Paris, France
R
Rian
3/19/2026 9:12:10 AM

This is very good and accurate. Explanation is very helpful even thou some are not 100% right but good enough to pass.

United States
G
Gerrard
3/18/2026 6:58:37 AM

The DP-900 exam can be tricky if you aren't familiar with Microsoft’s specific cloud terminology. I used the practice questions from free-braindumps.com and found them incredibly helpful. The site breaks down core data concepts and Azure services in a way that actually mirrors the real test. As a resutl I passed my exam.

United States
V
Vineet Kumar
3/6/2026 5:26:16 AM

interesting

Anonymous
J
Joe
1/20/2026 8:25:24 AM

Passed this exam 2 days ago. These questions are in the exam. You are safe to use them.

UNITED STATES
N
NJ
12/24/2025 10:39:07 AM

Helpful to test your preparedness before giving exam

Anonymous
A
Ashwini
12/17/2025 8:24:45 AM

Really helped

Anonymous
J
Jagadesh
12/16/2025 9:57:10 AM

Good explanation

INDIA
S
shobha
11/29/2025 2:19:59 AM

very helpful

INDIA
P
Pandithurai
11/12/2025 12:16:21 PM

Question 1, Ans is - Developer,Standard,Professional Direct and Premier

Anonymous
E
Einstein
11/8/2025 4:13:37 AM

Passed this exam in first appointment. Great resource and valid exam dump.

Anonymous
D
David
10/31/2025 4:06:16 PM

Today I wrote this exam and passed, i totally relay on this practice exam. The questions were very tough, these questions are valid and I encounter the same.

UNITED STATES
T
Thor
10/21/2025 5:16:29 AM

Anyone used this dump recently?

NEW ZEALAND
V
Vladimir
9/25/2025 9:11:14 AM

173 question is A not D

Anonymous
K
khaos
9/21/2025 7:07:26 AM

nice questions

Anonymous
AI Tutor 👋 I’m here to help!