Practice Test 3 | Google Cloud Certified Professional Data Engineer | Dumps | Mock Test
A high-rise mall building has a 5-story parking lot which visitors can use for free for the first hour, then with an hourly fee. Management is curious about what car brands usually use the parking lot in the building so they can predict the life quality of visitors to better distribute the different outlets within the mall’s stories, determine the best parking spots to stamp them as “VIP” for loyal customers.
To provide a better experience for customers and make it easier to collect data, the management looks for a solution to detect every car’s brand entering the parking lot. The management is considering the time and cost to build the detection model and looking for the outsource contractor who could achieve this the soonest. What would you do if you are asked to build the model?
A. Collect as many photos of cars entering the parking lot. Use AutoML Vision to build and train the model by using 70-80% of training photos you collected and use the rest of the training photos to test and tune the model.
B. Collect as many photos of cars entering the parking lot. Use AutoML Vision to build and train the model by using all the training photos you collected.
C. Collect as many photos of cars entering the parking lot. Use Dataproc to build the model using SparkML. Use 70-80% of training photos you collected to train the model and use the rest to train and tune the model. Deploy the model using Cloud ML Engine.
D. Collect as many photos of cars entering the parking lot. Use Cloud ML Engine with TensorFlow to build the model. Use all the training photos you collected to train the model. Deploy the model using Cloud ML Engine.
Correct Answer: A
Since you have a limited time and resources to build, train, and deploy the model, building your own model can be time-consuming and not in your favor. Google provides a great ML service called AutoML to quickly build models for you. AutoML Vision is one of its products which you can start with a training set as little as a dozen photo samples and AutoML takes care of the rest.
Option B is incorrect: AutoML Vision is the right choice. However, training the model with the whole training set is not the right approach in machine learning because you ought to test the model before considering it accurate enough for the production. Usually, the training set is split into 70-30% sets, first for training while the second one is for testing and tuning the model’s parameters.
Option C is incorrect: Using any approach other than AutoML can be time-consuming and with such a tight timeline, it’s not the best approach.
Option D is incorrect: Using this approach can also be time-consuming and using the whole training set for training is not the best practice as explained before.
Source(s):
Google Cloud AutoML:
https://cloud.google.com/automl/
Cloud Machine Learning Engine:
Comments are closed, but trackbacks and pingbacks are open.